U.S. Department of Commerce Volume 107 Number 2 April 2009 U.S. Department of Commerce Otto J. Wolff Acting Secretary of Commerce National Oceanic and Atmospheric Administration Jane Lubchenco, Ph.D. Administrator of NOAA National Marine Fisheries Service James W. Balsiger, Ph.D. Acting Assistant Administrator for Fisheries ^rEs o* Scientific Editor Richard D. Brodeur, Ph.D. Associate Editor Julie Scheurer National Marine Fisheries Service Northwest Fisheries Science Center 2030 S. Marine Science Dr. Newport, Oregon 97365-5296 Managing Editor Sharyn Matriotti National Marine Fisheries Service Scientific Publications Office 7600 Sand Point Way NE Seattle, Washington 98115-0070 The Fishery Bulletin (ISSN 0090-0656) is published quarterly by the Scientific Publications Office, National Marine Fisheries Service, NOAA, 7600 Sand Point Way NE, BIN C15700, Seattle, WA 98115-0070. Periodicals postage is paid at Seattle, WA. POSTMASTER: Send address changes for subscriptions to Fish- ery Bulletin, Superintendent of Docu- ments, Attn.: Chief, Mail List Branch, Mail Stop SSOM, Washington, DC 20402- 9373. Although the contents of this publica- tion have not been copyrighted and may be reprinted entirely, reference to source is appreciated. The Secretary of Commerce has deter- mined that the publication of this peri- odical is necessary according to law for the transaction of public business of this Department. Use of funds for printing of this periodical has been approved by the Director of the Office of Management and Budget. For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402. Subscrip- tion price per year: $36.00 domestic and $50.40 foreign. Cost per single issue: $21.00 domestic and $29.40 foreign. See back for order form. Editorial Committee John Carlson Kevin Craig Jeff Leis Rich McBride Rick Methot Adam Moles Frank Parrish Dave Somerton Ed Trippel Mary Yoklavich National Marine Fisheries Service, Panama City, Florida Florida State University, Tallahassee, Florida Australia Museum, Sydney, New South Wales, Australia National Marine Fisheries Service, Woods Hole, Massachusetts National Marine-Fisheries Service, Seattle, Washington National Marine Fisheries Service, Auke Bay, Alaska National Marine Fisheries Service, Honolulu, Hawaii National Marine Fisheries Service, Seattle, Washington Department of Fisheries and Oceans, St. Andrews, New Brunswick, Canada National Marine Fisheries Service, Santa Cruz, California Fishery Bulletin web site: www.fishbull.noaa.gov The Fishery Bulletin carries original research reports and technical notes on investigations in fishery science, engineering, and economics. It began as the Bulletin of the United States Fish Commission in 1881; it became the Bulletin of the Bureau of Fisheries in 1904 and the Fishery Bulletin of the Fish and Wildlife Service in 1941. Separates were issued as documents through volume 46; the last document was No. 1103. Beginning with volume 47 in 1931 and continuing through volume 62 in 1963, each separate appeared as a numbered bulletin. A new system began in 1963 with volume 63 in which papers are bound together in a single issue of the bulletin. Beginning with volume 70, number 1, January 1972, the Fishery Bulletin became a periodical, issued quarterly. In this form, it is available by subscription from the Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402. It is also available free in limited numbers to libraries, research institutions, State and Federal agencies, and in exchange for other scientific publications, U.S. Department of Commerce Seattle, Washington Volume 107 Number 2 April 2009 Fishery Bulletin Contents Articles Companion articles 109-132 Powell, Eric N., John M. Klinck, Kathryn A. Ashton-Alcox, and John N. Kraeuter Multiple stable reference points in oyster populations: biological relationships for the eastern oyster (Crassostreo virgimco) in Delaware Bay 133-147 Powell, Eric N., John M. Klinck, Kathryn A. Ashton-Alcox, and John N. Kraeuter Multiple stable reference points in oyster populations: implications for reference point-based management 148-164 The National Marine Fisheries Service (NMFS) does not approve, recommend, or endorse any proprie- tary product or proprietary material mentioned in this publication. No reference shall be made to NMFS, 165 — 174 or to this publication furnished by NMFS, in any advertising or sales promotion which would indicate or imply that NMFS approves, recom- mends, or endorses any proprietary product or proprietary material 1/3—103 mentioned herein, or which has as its purpose an intent to cause directly or indirectly the advertised product to be used or purchased because of this NMFS publication. 186-194 The NMFS Scientific Publications Office is not responsible for the con- tents of the articles or for the stan- dard of English used in them. Witthames, Peter R., Anders Thorsen, Hilario Murua, Francisco Saborido-Rey, Lorraine N. Greenwood, Rosario Dominguez, Maria Korta, and Olav S. Kjesbu Advances in methods for determining fecundity: application of the new methods to some marine fishes MacAvoy, Stephen E., Greg C. Garman, and Stephen A. Macko Anadromous fish as marine nutrient vectors Cartwright, Rachael L. Description of early life history stages of the northern sculpin ( Icelinus borealis Gilbert) (Teleostei:Cottidae) Ehrhardt, Nelson M., and Vallierre K. W. Deleveaux Management of fishing capacity in a spiny lobster (Panulirus argus) fishery: analysis of trap performance under the Florida spiny lobster Trap Certificate Program ii Fishery Bulletin 107(2) 195-206 Harter, Stacey l., Marta M. Ribera, Andrew N. Shepard, and John K. Reed Assessment of fish populations and habitat on Oculina Bank, a deep-sea coral marine protected area off eastern Florida 207-220 Sewall, Fletcher F., and Cara J. Rodgveller Changes in body composition and fatty acid profile during embryogenesis of quillback rockfish ( Sebastes maliger) 221-234 Nichol, Daniel G., and David A. Somerton Evidence of the selection of tidal streams by northern rock sole (Lepidopsetta polyxystra) for transport in the eastern Bering Sea 235-243 Graham, Larissa J., Mark L. Botton, David Hata, Robert E. Loveland, and Brian R. Murphy Prosomal-width-to-weight relationships in American horseshoe crabs (Limulus polyphemus): examining conversion factors used to estimate landings 244-260 Beacham, Terry D., John R. Candy, Khai D. Le, and Michael Wetklo Population structure of chum salmon (Oncorhynchus keta) across the Pacific Rim, determined from microsatellite analysis 261 Corrigendum Stark, James W. Geographic and seasonal variations in maturation and growth of female Pacific cod (Gadus macrocephalus) in the Gulf of Alaska and Bering Sea (Fish. Bull. 105t3]:404) 262 Guidelines for authors 109 Multiple stable reference points in oyster populations: biological relationships for the eastern oyster ( Crassostrea virgin tea) in Delaware Bay Eric N. Powell (contact author)' John M. Klinck2 Kathryn A. Ashton-AScox1 John N. Kraeuter1 Email address for contact author: eric@hsrl.rutgers.edu 1 Haskin Shellfish Research Laboratory Rutgers University 6959 Miller Avenue Port Norris, New Jersey 08349 2 Center for Coastal Physical Oceanography Crittenton Hall Old Dominion University Norfolk, Virginia 23529 Abstract — In the first of two com- panion papers, a 54-yr time series for the oyster population in the New Jersey waters of Delaware Bay was analyzed to develop biological relationships necessary to evaluate maximum sustainable yield (MSY) reference points and to consider how multiple stable points affect refer- ence point-based management. The time series encompassed two regime shifts, one circa 1970 that ushered in a 15-yr period of high abundance, and a second in 1985 that ushered in a 20-yr period of low abundance. The intervening and succeeding periods have the attributes of alternate stable states. The biological relationships between abundance, recruitment, and mortality were unusual in four ways. First, the broodstock-recruitment relationship at low abundance may have been driven more by the provi- sion of settlement sites for larvae by the adults than by fecundity. Second, the natural mortality rate was tem- porally unstable and bore a nonlin- ear relationship to abundance. Third, combined high abundance and low mortality, though likely requiring favorable environmental conditions, seemed also to be a self-reinforcing phenomenon. As a consequence, the abundance-mortality relationship exhibited both compensatory and depensatory components. Fourth, the geographic distribution of the stock was intertwined with abundance and mortality, such that interrelation- ships were functions both of spatial organization and inherent population processes. Manuscript submitted 29 November 2007. Manuscript accepted 9 September 2008. Fish. Bull. 107:109-132 (2009). The views and opinions expressed or implied in this article are those of the author and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. All federal fisheries, and some state fisheries, are managed under biologi- cal reference-point guidelines that implement a yearly allocation or quota, often termed TAC (total allow- able catch) or TAL (total allowable landing), to constrain fishing mortal- ity (e.g., Wallace et al., 1994). The biological reference-point approach for federal fisheries was mandated by the Magnuson-Stevens Fishery Conservation and Management Act (Anonymous, 1996) which requires management at a biomass that pro- vides maximum sustainable yield, Bmsy. Under this system, sophisticated survey, analytical, and modeling pro- cedures are used to identify selected biological reference points, such as the target biomass, BMSY, and carry- ing capacity, K. Fishing mortality is then set in relation to these goals. As a consequence, much attention has been given to the choice and applica- tion of biological reference points in fisheries management (e.g., Sissen- wine and Shepherd, 1987; Hilborn, 2002; Imeson et al., 2002; Mangel et al., 2002). Normally, BMSY is defined in rela- tion to carrying capacity, the biomass present without fishing, where natu- ral mortality balances recruitment (e.g., May et al., 1978; Johnson, 1994; Mangel and Tier, 1994; Rice, 2001). This stable point is characterized by a population in which most animals are adults, where natural mortality rates are low, and where recruitment is limited by compensatory processes such as resource limitation constrain- ing fecundity. BMSY is most commonly defined as — , based on the well-known Schaefer model which stipulates the guiding premise that surplus pro- duction is highest at ^ (Hilborn and Walters [1992]; see Restrepo et al. [1998] for more details on the feder- al management system; see NEFSC [19991, 20002, 20023] for examples of implementation of reference-point management). 1 NEFSC (Northeast Fisheries Science Center). 1999. 29th Northeast re- gional stock assessment workshop (29th SAW): Stock Assessment Review Com- mittee (SARC) consensus summary of assessments. NMFS NEFSC Ref. Doc. 99-14, 347 p. 2 2000. 30th Northeast regional stock assessment workshop (30th SAW): Stock Assessment Review Committee ( SARC ) con- sensus summary of assessments. NMFS NEFSC Ref. Doc. 00-03, 477 p. 3 2002. 34th Northeast regional stock assessment workshop (34th SAW): Stock Assessment Review Committee (SARC) con- sensus summary of assessments. NMFS NEFSC Ref. Doc. 02-06, 346 p. 110 Fishery Bulletin 107(2) Some have expressed concerns about managing at Bmsy (e.g., Peterman, 1977; Hilborn, 2002; Mangel et al., 2002), but only recently has the possibility been raised that carrying capacity may not be the long-term constant typically assumed under BMSY management. That realization arises ineluctably from the recognition that regime shifts profoundly affect the balance between population and environment (Rothschild, 2000; Collie et al., 2004; Rothschild and Shannon, 2004; Sakuramoto, 2005). Increasingly, fisheries biologists recognize these transitions as an important long-term component of population variation (e.g., Botsford, 1981; Steele and Henderson, 1984; Ware, 2000; Jackson et al., 2001; Choi et al., 2004; Collie et al., 2004; Breitburg and Fulford, 2006). Any change in carrying capacity assur- edly changes BMSY. The acceptance of regime shifts requires an acknowl- edgement that populations can exist in alternating sta- ble states that are self-reinforcing for protracted periods of time. The record of oyster abundance in Delaware Bay indicates at least two regime shifts (Powell et al., 2008), circa 1970 and circa 1985, with intervening and succeeding intervals having the attributes of alternate stable population states ( sensu Gray, 1977; Peterson, 1984; Knowlton, 2004). These periods of relative sta- bility are multigenerational and demonstrably not of anthropogenic origin4 (see Knowlton, 2004) because fishing mortality rates have been far below natural mortality rates over much of this time. The periods of stability are persistent over a range of climatic condi- tions (Soniat et al., in press). The association of unique climatic events with each of the regime shifts is consis- tent with models that emphasize the unique confluence of a set of forcing factors in the initiation of catastrophic events (DeAngelis and Waterhouse, 1987; Deakin, 1990; Hastings, 1991) and supports the observation of Collie et al. (2004) that large-scale changes in the population dynamics of species are commonly characterized by a poor correlation between the response variable and potential forcing factors. Evaluation of MSY-style reference points requires an understanding of the capacity of a species to expand its biomass over a range of biomasses. In fisheries parlance, this expansion capacity is related to surplus production. Regime shifts change expansion capacity in relation to biomass. Surplus production models are well described (e.g., Sissenwine and Shepherd, 1987; Maunder, 2003), but the influence of range shifts has rarely been con- sidered. In the first of two companion contributions, we develop relationships supporting a surplus production model for a species, the eastern oyster ( Crassostrea virginica), and a location, Delaware Bay, characterized by distinctive and well described range shifts. We take advantage of a 54-yr time series of oyster abundance, recruitment, and mortality for this analysis. 4 We recognize that the introduction of Haplosporidium nelsoni (MSX) circa 1957 (Burreson et al., 2000), which subsequently played a critical role in the 1985 regime shift, was likely anthropogenically driven. Table 1 The bed groups (by location: upbay and downbay) and subgroups (by mortality rate) for the eastern oyster ( Crassostrea virginica ) collected on twenty beds in Dela- ware Bay, as shown in Figure 1. Mortality rate divides each of the primary groups, themselves being divided by loca- tion, a surrogate for up bay-downbay variations in dredge efficiency and fishery area-management regulations. Bed group and subgroup Bed Upbay group Low mortality Round Island, Upper Arnolds, Arnolds Medium mortality Upper Middle, Middle, Sea Breeze, Cohansey, Ship John Downbay group Medium mortality Shell Rock High mortality Bennies Sand, Bennies, New Beds, Nantuxent Point, Hog Shoal, Hawk’s Nest, Strawberry, Vexton, Beadons, Egg Island, Ledge, Materials and methods The survey time series The New Jersey survey began as a response to overfish- ing that had reduced stock abundance by the early 1950s. By 2006, this 54-yr record covered a number of unique periods, including the period of time after the onset of MSX, a disease caused by the protozoan Haplosporidium nelsoni, circa 1957 (Haskin and Andrews, 1988; Ford, 1997) and the period after the onset of Dermo, a disease caused by the protozoan Perkinsus marinus, circa 1990 (Ford, 1996; Cook et al., 1998). In what follows, we define the population on the twen- ty primary oyster beds in Delaware Bay (Fig. 1) as the oyster stock in the New Jersey waters of Delaware Bay, but for simplicity we refer to it as the Delaware Bay oyster stock.5 The analyses that follow will delineate four bed groups based on the long-term average rates of natural mortality, productivity, and survey catchability (Table 1). Analyses of the Delaware Bay oyster resource 5 Oysters are also found on the Delaware side of the bay, although the total bed area is much less than that in the New Jersey waters (Moore, 1911; Maurer et al., 1971; Maurer and Watling, 1973), as well as in many of the river mouths; and an unknown number (but significant during certain periods of history LMacKenzie, 1996; Ford, 1997]) have been present on leased grounds, most of which are situated downbay of Egg Island (see Fig. 1 of Haskin and Ford, 1982). Inadequate survey data exist to include oysters in bay margin habi- tats and on leased grounds in the stock analysis. Delaware maintains an independent survey, but these data are not yet available on a per-m2 basis. However, abundance and recruitment trends typically have been similar on both sides of the bay. Powell et at: Multiple stable reference points in oyster populations: Crassostrea virgmica in Delaware Bay 111 75° 30' 75° 20' 75° 10' 39° 25' 39° 20' 39° 15' 39° 10' 39° 25' 39° 20' 39° 15' 39° 10' 75° 30' 75° 20' 75° 1 0' Figure 1 The twenty natural oyster beds of the eastern oyster ( Crassostrea virginica ) in the New Jersey waters of Delaware Bay may be characterized in terms of high-quality (dark shade) and medium-quality (light shade) grids, the term quality referring to a relative differential in long-term average oyster abundance (Powell et al, 2008). The footprints for the Middle bed (upper portion of figure) and the beds downbay from it, excepting New Beds, Egg Island, and Ledge, were updated with data from surveys in 2005 and 2006. The footprints for the remaining beds were based on historical definitions. of New Jersey routinely reveal a division between an upbay group of eight beds (Round Island, Upper Ar- nolds, Arnolds, Upper Middle, Middle, Sea Breeze, Co- hansey, and Ship John) and a downbay group of twelve beds (Shell Rock, Bennies Sand, Bennies, New Beds, Nantuxent Point, Hog Shoal, Hawk’s Nest, Strawberry, Vexton, Beadons, Egg Island, and Ledge) (Fig.l). Salin- ity, natural mortality rate, and growth rate are higher downbay. Dredge efficiencies are significantly higher downbay (Powell et al., 2002a, 2007). Both regions can be subdivided by natural mortality rate and productiv- ity. In the upbay group, natural mortality rates and growth rates are significantly lower for the upper three beds, Round Island, Upper Arnolds, and Arnolds, than for the remaining beds. Henceforth these two groups will be termed “the low-mortality” and “medium-mor- tality” beds, respectively (Table 1). In the downbay group, growth rates and mortality rates are lower for 112 Fishery Bulletin 107(2) Figure 2 Time series of abundance of the eastern oyster ( Crassostrea virginica) in Delaware Bay, showing four subgroups defined by location and natural mor- tality rate. Total oyster abundance for any year is the sum of abundance in the subgroups. Beds in the subgroups are listed in Table 1. Figure 3 Time series of spat recruitment per >20-mm eastern oyster ( Crassostrea vir- ginica) in Delaware Bay. Solid and dashed lines mark the 1.0 and 0.5 spat- to-oyster levels, respectively. Shell Rock, leading to its designa- tion as a medium-mortality bed; the remainder are high-mortality beds (Table 1). Powell et al. (2008) have de- scribed the Delaware Bay time series in detail. The pertinent findings are summarized in the following sections. Pre-1970 period of low abundance In the few years before 1957 when survey data were available, the Delaware Bay oyster population was characterized by relatively low abundance (Fig. 2), an unre- markable rate of recruitment (Fig. 3), relatively low natural mortality (Fig. 4), and a spatial distribution in which the fraction of the stock on the medium-mortality beds was rel- atively low in comparison with the 54-yr median of 0.417 (Fig. 5). The dispersion of the stock was likely maintained by overfishing because the fishery predominantly targeted the medium-mortality beds during this time (Powell et al., 2008). Given that natural mortality rates averaged below 10% during this period, and fishing rates routinely exceeded 10%, we speculate that, had fishing rates been the same as those in later years (typically <7% of the stock), the medium-mortality beds likely would have contributed a larger proportion of the stock, and stock abundance likely would have been higher than that observed. MSX entered the picture circa 1957. Abundance was unchanged, in part because of implementation of reference point-based manage- ment that curtailed overfishing (Fegley et al., 2003; Powell et al., 2008). The early reference point referred to as “the 40% rule” lim- ited removals from individual beds when the volume of live oysters de- clined below 40% of a bushel haul (Powell et al., 2008). The 40% rule successfully limited harvest from the late 1950s until the 1985 re- gime shift, after which changes in the fishery imposed by low abun- dance and Dermo required develop- ment of management alternatives and new reference points (Powell et al., 2008). Under the 40% rule, Powell et al. : Multiple stable reference points in oyster populations: Crassostrea virgimca in Delaware Bay 113 the highest fishing mortality rate observed after 1958 was about 10% of the stock (Powell et al., 2008). By circa 1960, the effect of an increase in natural mortal- ity, on the order of 5-10% of the stock, had been ameliorated by a decrease in fishing mortality at least that large. From 1957 through 1966, natural mortality neared 15% of the stock in most years and exceeded 20% in two years (Fig. 4). Mortality substan- tively increased downbay and by 1960, animals on the high-mor- tality beds were contributing a disproportionate share of the to- tal mortality of the population (Fig. 6). As a consequence, during the 1960s, individuals on the me- dium-mortality beds contributed more than their long-term me- dian proportion of the total stock in eight of ten years (Fig. 5). Al- though the fishery continued to target these beds (Powell et al., 2008), the reduction in total re- movals minimized the influence of the fishery on the stock. The 1970 population expansion In 1970, the oyster population increased by more than a factor of two, and this high level of abundance was maintained for the succeeding 15 years. This was a period of high abundance in a number of other species of commercial importance (Gabriel, 1992; Link et al., 2002), includ- ing many finfish species in the Gulf of Maine and Mid-Atlan- tic Bight, hard clams along the Long Island coast (Kraeuter et al., 2005; Hofmann et al., 2006), and Illex squid off Newfoundland (Dawe et al., 2000). In many of these cases, this abundance was rapidly impacted by overfishing (e.g., Kraeuter et al., 2008), which artificially limited its duration. A decline in population, however, did not occur for the Delaware Bay oyster stock. However, the general coincidence of abundance in bay and shelf species, both temperate and boreal, bespeaks of a large-scale climatic event that influenced much of the northeast Figure 4 The fraction of eastern oyster ( Crassostrea virginica ) dying each year in the New Jersey waters of Delaware Bay, 1953-2006.. Horizontal line marks an arbitrary boundary between mortality in epizootic (above the line) and non- epizootic years. Year Figure 5 The fraction of the total stock of eastern oyster ( Crassostrea virginica) in the New Jersey waters of Delaware Bay that was located on the medium-mortality beds, 1953-2006. The horizontal bar represents the 54-yr median of 0.417. 114 Fishery Bulletin 107(2) 0.8- 75 0.6-| o 0.5- 0.3- 0.1- I I III I I III III 1111 I I I I I I I I I I I I I minmincocococDco cococooococDcncncncn 0)0)0)0)0)0)0)010)0)0)0)0)0)0)0)0)0)0)0)0)0)0)0) t— CM CM CM Year Figure 6 The fraction of total deaths in the population of eastern oyster ( Crassostrea virginica ) in the Delaware Bay contributed by the individuals on the high-mor- tality beds. The horizontal bar represents the 54-yr median level of 0.482. U.S. coastline. Baines and Folland (2007) documented the climatic forcing that certainly provided the basis for the 1970 regime shift, although how climate change in the North Atlantic imposed the conditions for increased productivity on the local scale remains uncertain. Two noteworthy events preceded population expan- sion in Delaware Bay. First, 1968-70 were three suc- cessive years of relatively high recruitment (Fig. 3). Only one other trio of such years, 1997-99, exists in the time series. Relatively high recruitment in these three years occurred in three of four bay subgroups (medium-mortality, Shell Rock, and high-mortality). No equivalent coincidence of years and bay coverage exists in the time series. Second, beginning in 1967, natural mortality dropped below 10% after the largest MSX epizootic event of the 1960s and remained at or below this level through 1975 (Fig. 4). The coincidence of dramatically lower natural mortality and a triplex of high recruitment years was unique in the time series and certainly provided the proximate conditions for the population expansion of 1970. The 1970-85 high-abundance interval and its termination The 1970-85 time period was remarkable for its per- sistent high level of oyster abundance (Fig. 2). The period was characterized by a lower contribution of animals on the high-mortality beds to total population mortality (Fig. 6) and by natural mortalities that rarely exceeded 13% of the stock annually (Fig. 4). During this period, the fraction of deaths on the high-mortality beds exceeded the long-term median only six times (Fig. 6). In the first half of the period, the medium-mortality beds contributed proportionately more to the stock, as they had during most of the MSX-dominated decade that preceded this period (Fig. 5). High freshwater inflow contributed to sustainable high abundance by limiting mortality from MSX. A dramatic shift in stock disper- sion began in 1979, coincident with the cessation of consistently high freshwater inflows, and led, over a few years, to proportional increases in abundance in the more environmentally sensitive waters of the upbay and downbay margins. An increase in the susceptibility of the population to epizootic disease mortality consequent of the increased abundance downbay evolved from 1985 to 1986 through a coincidence of climatic events into the largest epizootic event in the recorded history of Delaware Bay (Fig. 4). Interestingly, the 1985-86 stock collapse was not obviously associated with any unusual trends in recruitment immediately before or after the collapse (Fig. 3), nor did the distribution of deaths (Fig. 6) or the dispersion of the stock (Fig. 5) change. Abun- dance declined in all bay regions. The post-MSX period The few years immediately following the 1985-86 MSX epizootic event and preceding the onset of Dermo circa 1990 were not unusual in any way, and neither was the first half-decade after Dermo became an impor- tant contributor to population mortality. Total abun- Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virgmica in Delaware Bay 115 dance remained relatively stable from 1987 through 2001 (Fig. 2). Recruitment was not unusual (Fig. 3). However, natural mortality rose dramatically, from the 10% level immediately after 1986, to often exceed 20-30% throughout the 1990s (Fig. 4). The fraction of deaths contributed by the high-mortality beds did not change markedly over the 1990s, although the fractions of deaths did rise incrementally in 1990 compared to the few preceding years (Fig. 6). The dispersal pattern of the 1980s remained through 1995 (Fig. 5), despite the increased mortality rate on the high-mortality beds. The response of the stock to Dermo became more ap- parent in 1996, when the stock began a rapid contraction to its refuge on the medium-mortality beds. This con- traction in dispersion occurred at the same time as in- creased recruitment on these beds (Fig. 5) and counter- weighed the accumulating losses of individuals farther downbay (Fig. 6), so that total abundance did not change. The post-2000 era Although the time series is still limited in scope, a change in population dynamics is evident around 2000. Beginning in 2000, the recruitment rate declined pre- cipitously and remained low at least through 2006 (Fig. 3). Total abundance declined with continuing high mor- tality on the high-mortality beds (Fig. 6), but stock consolidation continued, with an increasing proportion of animals on the medium-mortality beds. As a conse- quence, mortality in the population as a whole declined (Fig. 4). The fraction of total mortality contributed by the high-mortality beds declined to its lowest level since the 1950s and remained low (Fig. 6) because consolida- tion of the stock upbay limited the number of individuals available to die on the high-mortality beds. Overview of fishing activities The analysis that follows makes reference to two distinc- tive types of fishing on the Delaware Bay oyster beds of New Jersey. From 1953 through 1995, a “bay-season” fishery occurred, in which a portion of the beds was opened, usually for 2-6 weeks in the spring. Oysters were removed en masse and transplanted downbay to leased grounds. Based on recent dredge efficiency esti- mates (Powell et al., 2007), the method for transplanting was relatively nonselective for oyster size; oysters were moved more or less in proportion to their contribution to the size-frequency distribution of the population. In most years, the fishery was limited by the 40% rule. As a consequence, target beds varied during the pro- gram from year to year as the relative abundance of the resource varied. Since 1996, a direct-market fishery has been pros- ecuted for the most part on beds from Shell Rock down- bay (Fig. 1). In this fishery, market-size oysters are taken directly off the beds and marketed immediately or stored for a time on leased grounds before they are marketed. The vast majority of animals removed by this fishery have exceeded 63 mm (Powell et al., 2005). Model formulations and statistics Basic population dynamics Quantification of the Dela- ware Bay time series has been described in Powell et al. (2008). Natural mortality fractions were obtained from box counts under the assumption that N — N, + N, ( 1 ) oysters boxest Live oysters ^ > ’ where N = the number of individuals; and t - any given year. Hence N, <*> bc = boxesf Ni + Ni boxesf Live oysters f (2) where 0, was obtained by difference: 0>o = t Nt - Nt_, ) - ( Rt_, - - a>fNt^ ; Nt-i + Rt-i (3) where 0, varied randomly over the time series, with a 54-yr mean of 0.274 and a 54-yr median of 0.311 (Powell et al., 2008). 116 Fishery Bulletin 107(2) Forces modifying abundance: broodstock-recruilment relationship A linear fit to the broodstock and recruit- ment data returned a regression coefficient of only 0.076 (Fig. 7). The relationship was strongly compen- satory. A variety of broodstock-recruitment models might be applied (e.g., May et ah, 1978; Hilborn and Walters, 1992; Kraeuter et al., 2005), given the scatter of data at high abundance and the paucity of extremely high values. We used a relationship that produced declining recruitment at high abundance (overcom- pensation sensu Hancock, 1973; McCann et al., 2003), because shellfish can achieve densities sufficient to limit growth and reproduction (e.g., Frechette and Bourget, 1985; Frechette and Lefaivre, 1990; Powell et al., 1995). Application of the simple filtration model of Wilson-Ormond et al. (1997) indicated that present- day abundances, even on the medium-mortality beds, are below such densities, but abundances in the 1970s were very likely high enough and medium-mortality abundances circa 2002 (Fig. 2) may have been high enough to restrict growth. Thus, from Hilborn and Walters (1992): where R = the number of spat in millions; and N, oyster abundance in millions. Fitting this curve to the data for the high- and medium- quality strata (Fig. 1) yields a - 0.3746, and /3 = 5121.9 (Figs. 7 and 8). We compared the result of Equation 4 to the result of a best-fit linear regression with zero intercept (Fig. 8). The linear relationship is Rt =0.493 Nt. (5) Rt = Nt_lt a l+; U-l (4) Broodstock and box-count mortality Box-count esti- mates of natural mortality are also related to trends in abundance (Fig. 9). At abundances greater than 4 x 109, mortality was low. The fraction dying each year averaged 9.6% for these nonepizootic years, defined for convenience as years in which the fraction dying was less than 20%. The nonepizootic death rate was rela- tively independent of abundance, although the lowest mortalities, less than 6%, occurred at abundances below 6 x 109. Of the 14 epizootic years in the 54-yr record, defined in our study as deaths exceeding 20% of the stock, 13 occurred at abundances less than 3xl09 (Fig. 9). The exception was 1985. Of the 32 years with abundances less than 3 x 109, 14 were epizootic years. Of these 32, only one had a fractional mortality be- tween 0.15 and 0.20. Accordingly, two divergent outcomes existed over a range of low abundances. In some years, the fraction dying approximated the long- term mean for high-abundance years, about 9.6%. In other years, epizootic mortalities occurred. Epizootic events also occur rarely at very low abundanc- es. Note on Figure 9 that no mortality fraction exceeded 0.17 at abundances below 1.5 xlO9. Thus, a complex rela- tionship exists between abundance and mortality. We apply an admittedly ad hoc em- pirically derived equation to describe the relationship between box-count mortality and abundance depicted in Figure 9: Figure 7 The broodstock-recruitment relationships for eastern oyster (Crassostrea virginica), 1953-2006. The solid line is the best-fitted Ricker curve (Eq. 4). The dashed line is a second-order polynomial fit (see Kraeuter et al., 2005). Note that the polynomial fit overestimates recruitment at high abundance. Dotted lines (vertical and horizontal) mark the 54-yr medians of abundance and recruitment. =co + K\oge(Nt_1 + p ) -(pN^ + xN^e Nf_ i- t-i-v 2v (6) where to = 0.055; k = 0.03; P = 1.0; cp = 0.0025 X - 0.1; xp = 2.2; v = 0.8; and N is expressed as billions of animals. Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virginica in Delaware Bay 117 70x109- 6.0x109- 0.0x10°' 0.0x10° 5.0x10s 1 .0x1 09 1.5x109 2.0x109 2.5x109 3.0x109 3.5x109 4.0x109 4.5x109 5.0x109 Oyster (>20 mm) abundance Figure 8 The low-abundance portion of the broodstock-recruitment relationship for the natural oyster beds of eastern oyster ( Crassostrea virginica ) in Delaware Bay, 1953-2006. The dotted line is the best-fitted Ricker curve (Eq. 4), also shown in Figure 7 for the entire data set. The solid line is a linear fit (Eq. 5) with zero intercept. Note that at low abundance, the linear fit travels through the recruitment values slightly below that traversed by the Ricker curve. Equation 6 has the unique property of eliciting both depensatory and compensatory trends at low abun- dance. Sissenwine (1984), Hilborn and Walters (1992), and Peterson et al. (2001) have provided exam- ples of the well-known depensatory process in which increased preda- tory mortality rate is associated with increased prey population density because of increased prey preference at high prey density. Allen (1979) provided a somewhat unusual case for depensation in oysters determined by substrate availability rather than by disease. Hilborn and Walters (1992) pro- vided an analogous example from human exploitation of declining fish stocks. The present case is unusual, however, because box- count mortality first increases with declining abundance, but this depensatory phase is then followed by compensation in the mortality rate as abundance continues to decline. Calculation of first passage time Mean first-passage times were calculated from Redner (2001), according to the methods of Roth- schild et al. (2005) and Rothschild and Mullen (1985). Input data were obtained by divid- ing a two-dimensional data set into quadrants by the medians of the x and y variables (Fig. 10). An example frequency table for the broodstock and recruitment relationship (Table 2) shows the frequency of occur- rence of the data from the 54-yr time series in each of the four quadrants, employing the quadrant number- ing convention depicted in Figure 10. For instance, years characterized by low abundance and low recruit- ment, thus falling into quadrant 1, occurred 32% of the time. Table 2 also displays one-year transition probabilities compiled by examining the quadrant location of the x-y datum in successive years. For example, a low-recruitment -t-low-abundance year fall- ing into quadrant 1, was followed one year later by a high-recruitment+high-abundance year, an occurrence falling into quadrant 4, 18.8% of the time, whereas 50% of the time, the following year was also a low- recruitment+low-abundance year. Thus, given that quadrant 1 is the starting point, the interval of time in which the population finds itself back in quadrant 1 should be a lesser number of years than the time required for the population to shift from quadrant 1 to quadrant 4. Mean first passage times (Table 3) express the number of years likely to elapse before the population with the x-y relationship characteristic of any one quadrant is again described by the relation- ship characteristic of that same quadrant, or obtains the relationship characteristic of one of the three other quadrants. Results and discussion Biological relationships that determine population dynamics Broodstock and recruitment A relationship between broodstock and recruitment is commonly found for shell- fish (Hancock, 1973; Peterson and Summerson, 1992; McGarvey et al., 1993; Lipcius and Stockhausen, 2002; Kraeuter et al., 2005), although not in every case has one been observed (Hancock, 1973; Crocos, 1991; Honkoop et al., 1998; Livingston et al., 2000). Such a relationship is commonly assumed for population dynamics models, and the adequacy of these models supports the likely importance of such a relationship in oysters (Mann and Evans, 1998, 2004; Dekshenieks et al., 2000; Powell et al., 2003). However, empirical evidence in oysters is contradictory and not well documented (e.g., Hofstetter, 1983; Mann et al., 1994; Southworth and Mann, 1998; Livingston et al., 2000), and the travails of larval life and at settlement are certainly likely to add consider- able uncertainty to the success of any search for such 118 Fishery Bulletin 107(2) evidence (Osman et al., 1989; Powell et al., 2002b, 2004; Hofmann et al., 2004). In Delaware Bay, recruitment rates below 2 x 109 spat are disproportionately associated with abundances of less than 3xl09 oysters (Fig. 7). The distribution of years in the four quadrants of the broodstock-recruit- ment diagram was 17, 9, 9, and 18 for quadrants 1, 2, 3, and 4 (as defined in Fig. 10), respectively (Table 2). This distribution was unlikely by chance, given the expecta- tion that one-quarter of the years should fall into each quadrant: P-0.10, P<0.10; P<0.10; P<0.10, for quadrants 1-4, respectively (binomial test: p = 0.25, q = 0.75). Twice as many high-recruitment events were associated with high abundance than with low abundance, and about twice as many low-recruitment events were associated with low abundance than with high abundance. The 54- yr average recruitment rate, expressed as the number of spat per >20-mm oyster per year, was 0.959. The median was lower, at 0.600. The long-term likelihood of a one-year population-replacement event (i.e. one spat per >20-mm oyster) was 17 in 54, and a recruitment rate half that high occurred in 27 of 54 years (Fig. 3). Only four massive recruitment events (>1.7 x 109 spat) occurred over the 54 years (Fig. 7). The rarity of these occurrences is not unusual (e.g., Loosanoff, 1966; Hofstet- ter, 1983; Oviatt, 2004; Southworth and Mann, 2004). The events were not predict- ed by the broodstock-recruitment curve. In most years, however, the broodstock-re- cruitment relationship was relatively pre- dictive, and the vast majority of recruits sustaining the population over the 54 years accrued from the 50 more-standard recruit- ment events. Nevertheless, even in average recruitment years, variability about the curve was large, about 4xl09 spat. Mean first-passage times calculated from one-year transition probabilities (Table 2) varied from 3 to 8 years (Table 3). Return intervals were about 3 years for a popula- tion beginning in quadrant 1 (low recruit- ment and low abundance) returning to quadrant 1, and for a population beginning in quadrant 4 (high recruitment and high abundance) returning to quadrant 4. The longest return intervals were associated with quadrant 3 (low recruitment and high abundance) as a destination. A population beginning in quadrant 2 or quadrant 3 was somewhat more likely to fall to quadrant 1 than to move to quadrant 4. Thus, overall, populations at low abundance were likely to remain there (quadrant 1) because of low recruitment, whereas populations at high abundance were likely to remain there because of high recruitment. Quadrants 1 and 4 have the characteristics expected of stable states. The broodstock-recruitment relation- ship (Fig. 7) indicated that the number of recruits per adult declined at high abun- dance. Note in particular (Fig. 3) that the number of recruits per adult was not unusually high during the 1970-85 high- abundance period, with the exception of 1972. In fact the number of one-year re- placement events (i.e. one spat per adult) was lower for a longer time during this Figure 9 The relationship between oyster abundance and box-count mortality for the eastern oyster ( Crassostrea virginica ) on the natural oyster beds of Delaware Bay during 1953-2006. The solid line is the curve described by Equation 6 fitted to the data. Dotted lines mark the 54-yr medians of abundance and box-count mortality (Table 4). (A) represents entire data set; (B) focuses on the region of the diagram encompassing abundances below 5 x 109, an abundance range in which an increase in mortality frequently occurs due to epizootic events. Years characteristic of regime shifts are identified in A only. Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virginica in Delaware Bay 119 Table 2 One-year transition probabilities and the frequency of occurrences for the eastern oyster ( Crassostrea virginica) population in each quadrant over the 54-yr time series were calculated from the Dela- ware Bay oyster broodstock-recruitment distribu- tion (Fig. 7). Median abundance over 54 years was 2.64xl09 and median recruitment was 1.53xl09. Arrows indicate trajectories between quadrants. Quadrants are defined in Figure 10. Quadrant 1 2 3 4 1 0.500 0.125 0.188 0.188 2 — > 0.444 0.222 0.000 0.333 3 0.222 0.333 0.333 0.111 4 — * 0.111 0.111 0.167 0.611 Frequency of occurrence 0.320 0.170 0.170 0.340 Number of years 17 9 9 18 15-yr period than at any other time before 2000. Thus, high broodstock abundance was not reward- ed by equivalently high recruitment. Three mech- anisms seem viable. The first is that fecundity declines at high abundance as availability of food becomes limited. Food limitation by high densities of filter feeders is well described (e.g., Peterson and Black, 1987; Rheault and Rice, 1996; Wilson- Ormond et al., 1997). The second is that cannibalism of larvae occurs, but this cause of mortality is of unlikely importance (Andre et al., 1993; Tamburri et al., 2007). The third is that predation rates on juveniles increase at high abundance. Although little evidence of this ef- fect exists (e.g., Whitlatch and Osman, 1994; Powell et al., 1995), a proportional increase in predation on juveniles at high abundance is consistent with optimal foraging theory (Hughes, 1980), under the assumption that oyster predators are optimal foragers (Powell et al., 1995; see also Pyke, 1984; Pierce and Ollason, 1987). All are standard explanations for compensation in the broodstock-recruitment relationship (e.g., Myers and Barrowman, 1996). The broodstock-recruitment diagram (Fig. 7) indi- cates that low abundance limited total recruitment in some way. This relationship is clear despite the exclu- sion from this data series of an unknown number of adults and recruits in State of Delaware waters, along the fringes of the bay, particularly in the river mouths, and on the leased grounds downbay of the high-mortal- ity beds. Moreover, the leased grounds likely retained substantial numbers of adult animals before the mid- 1980s, although estimates of abundance are not avail- able. Many fewer were present thereafter because of the demise of the bay-season fishery.6 Interestingly, the >- .Q CO Quad L rant 2 Quadrant 4 Quad rant 1 Quadrant 3 Variable X Figure 10 Mean first passage times for eastern oyster ( Crassostrea vir- ginica) were calculated by employing an arbitrary quadrant numbering convention. One-year transition probabilities were obtained by examining the position of consecutive x-y data pairs in quadrant space. Four transitions are possible for each starting position, the possibilities for quadrant 1 being depicted. Sixteen total trajectories are possible. Table 3 Mean first passage times, as well as the distribution of occurrences of the eastern oyster ( Crassostrea virginica ) population in each quadrant, after an infinite number of steps were calculated from the Delaware Bay oyster brood- stock-recruitment distribution (Fig. 7). The observed distribution of occurrences is given in Table 2. Arrows indicate trajectories between quadrants. Quadrants are defined in Figure 10. Quadrant 1 2 3 4 Mean first passage time (yr) 1 — > 3.25 6.06 6.45 5.00 2 — > 3.60 5.78 7.82 4.14 3 -> 4.20 4.56 5.78 5.24 4 — > 5.40 6.26 6.65 2.89 Distribution after an infinite number of steps 0.308 0.173 0.173 0.346 6 Anecdotal information indicates that numbers were low in the 1960s as well. decline in abundance on leased grounds after 1985 does not generate a perceptible change in the broodstock-re- cruitment relationship. Oyster larvae tend to set preferentially on live oysters and boxes rather than on cultch (shell clumps, shells, 120 Fishery Bulletin 107(2) and shell fragments without attached live oysters or boxes) (Powell et al., 2008); therefore, one possible ex- planation for the relationship between broodstock and recruitment is that adult abundance increased settle- ment success by providing a principal source of clean shell. Two avenues of evidence support this idea. First, a recruitment-enhancement program initiated in 2005 strongly indicated that Delaware Bay is not larvae-lim- ited, even at low population abundance levels (unpubl. data, first author). Clean shell planted at the appropri- ate time consistently sustains a settlement rate 5 to 10 times that for native shell. Second, the relationship be- tween adult numbers and recruitment held for the bay overall, even though the numbers of animals in various regions of the bay varied relatively independently, and independently of numbers for the bay as a whole. The broodstock-recruitment relationship was nearly identi- cal for two key bay areas, the medium-mortality and high-mortality beds (Fig. 11), despite widely and inde- pendently varying abundances over the time series (Fig. 5). Different trajectories would have been expected if recruitment rate depended upon a stock-wide abundance with trends divergent from local peregrinations. Broodstock and mortality Epizootics, here defined as bay-wide disease-induced mortality events affecting greater than 20% of the stock, occurred in about half of the years since 1989 (Figs. 4 and 9), but with much lower frequency in prior years. Deaths in nonepizootic years affected on average around 10% of the stock. All but one of the epizootics occurred at abundances between 1.5 x 109 and 4 x 109. The single outlier occurred at just over 10 x 109 animals; this is the 1985 MSX epizo- otic event that terminated the high-abundance period of the 1970s. The remaining events included the relatively few MSX epizootics of the 1950s and 1960s and the more frequent Dermo epizootics of the 1990s and 2000s. The distribution of data points in the four quadrants based on information in Figure 9 was 9, 17, 17, and 10 in quadrants 1, 2, 3, and 4, respectively (Table 4). This distribution is unlikely to occur by chance, but barely so: P<0.10, P-0.10; P-0.10; P>0.10, for quadrants 1-4, respectively (binomial test: p = 0.25, q- 0.75). Note that the use of the median mortality of 0.127 to define highl- and low-mortality quadrant groups yields a number of nonepizootic years in the same quadrants as the epizootic years (those with mortalities exceeding 0.20). Thus, the high-mortality quadrants include years when mortalities were not extraordinarily high. Note also that high-abundance years, those with abundance exceeding the median of 2.64 xlO9, include a few epizootic years with abundances near the median. That is, the use of Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virginica in Delaware Bay 121 Table 4 One-year transition probabilities, as well as the fre- quency of occurrence, of the eastern oyster ( Crassostrea virginica) population in each quadrant over the 54-yr time series were calculated from the Delaware Bay oyster broodstock-mortality distribution (Fig. 9). Median abun- dance was 2.64xl09 and the median mortality fraction was 0.127. Arrows indicate trajectories between quad- rants. Quadrants are defined in Figure 10. Quadrant 1 2 3 4 1 — > 0.222 0.444 0.222 0.111 2 — > 0.125 0.500 0.063 0.313 3 — > 0.059 0.059 0.647 0.233 4 — > 0.300 0.400 0.300 0.000 Frequency of occurrence 0.170 0.320 0.320 0.189 Number of years 9 17 17 10 medians allocates most, but not all, epizootic years in the abundance range of 1.5xl09 to 3 x 109 to a single quadrant. Nevertheless, even with this ambiguity, high-mortal- ity events were more likely with low abundance and some transitions were more likely to occur than oth- ers. Mean first-passage times were particularly long for transitions to quadrant 1 (low-mortality+low-abun- dance), always exceeding 6 years (Table 5). Mean first- passage times were also long for most transitions to quadrant 3, the low-mortality+high-abundance quad- rant, with the exception of those with quadrant 3 as the initial state. By contrast, the population was likely to return to quadrant 2 (high-mortality+low-abundance) from most quadrants in about 3-4 years (Table 5). This return interval is an expression of the relative fre- quency of Dermo epizootics. Interestingly, the tendency to return to quadrant 2 (high-mortality+low-abundance) was distinctly less from quadrant 3 (low-mortality and high abundance) than from other quadrants. High-mor- tality events were unlikely to occur when abundance was high. The distribution of first-passage times again supports the presence of multiple stable states for the Delaware Bay oyster population. The distribution of mortality with abundance is not constant, nor does it display a simple density depen- dency. Epizootics occurred less often at high abundance and near lowest abundance. Decreased mortality at low abundance was not unexpected for a population exposed to a disease that generates epizootic conditions (Gill, 1928; Ackerman et al., 1984; Kermack and McKend- rick, 1991). Normally, transmission rates of disease decline with decreased host density because contact rates decrease (Black, 1966; Andreasen and Pugliese, 1995; Godfray and Briggs, 1995; Heesterbeek and Rob- erts, 1995) and this leads to lower rates of mortality. This decline in transmission rates is true for nearly all diseases but does not seem to be the case for MSX Table 5 Mean first passage times as well as the distribution of occurrences of the eastern oyster ( Crassostrea virginica) population in each quadrant after an infinite number of steps were calculated from the Delaware Bay oyster broodstock-mortality distribution (Fig. 9). The observed distribution of occurrences is given in Table 4. Arrows indicate trajectories between quadrants. Quadrants are defined in Figure 10. Quadrant 1 2 3 4 Mean first passage time (yr) 1 — > 6.54 3.55 6.14 4.59 2 6.87 3.03 7.09 3.67 3 — > 8.10 6.00 3.11 4.21 4 — * 6.18 3.88 5.68 5.11 Distribution after infinite steps 0.153 0.339 0.321 0.196 or Dermo, which are characterized by inherently high transmission rates over a wide range of abundance (Hofmann et al., 1995; Powell et al., 1996, 1999). In the Delaware Bay oyster stock, the declining frequency of epizootics at low abundance originates in the dynam- ics of stock dispersion. A contraction of the stock away from areas of highest disease mortality normally is as- sociated with low abundance. Thus, epizootics are most likely to occur in a narrow window of abundance as the stock expands from its habitat of refuge on the medium- mortality beds, thereby leaving a greater proportion of the stock once again on the medium-mortality beds. This stock contraction, consequently, mitigates against a recurrence of the high-mortality event. Depensation in the mortality rate as abundance declines is, of course, an extinction scenario, were it to continue. The coun- tervailing compensatory process of stock contraction is the dominant protective action against local extinction, rather than a decline in host density that reduces dis- ease transmission rates. What is unusual is the low probability of epizootics at high abundance. Mortality rates are often assumed to be invariant over a wide abundance range for marine species (e.g., Paloheimo, 1980; Hoenig, 1983; Vetter, 1987; Clark, 1999) and, contrariwise, increased mor- tality at high abundance is expected of most popula- tions exposed to epizootic disease (e.g., Anderson and Gordon, 1982; Andreasen and Pugliese, 1995; Godfray and Briggs, 1995; Jaenike, 1998). Neither expectation conforms to what has been observed. Thus, one of the interesting quandaries is the maintenance of popula- tion abundance near the higher carrying capacity of the population during the 1970s-1985 high-abundance period. Some portion of this was caused by reference point-based management, which controlled fishing mor- tality to values normally below 5% of the stock (Pow- ell et al., 2008). Some portion was due to higher than 122 Fishery Bulletin 107(2) average freshwater inflows for much of the 1970s, which limited the influence of MSX. However, the fact that high abundance continued for at least five years after freshwater inflows subsided to more normal conditions circa 1979, and the depensation in the abundance-mor- tality relationship, would indicate that high abundance may reduce the probability of epizootics. This possibility has been treated theoretically by Powell et al. (1996), who showed that simulated oyster populations under- going significant increases in abundance were very unlikely to also generate Dermo epizootics. Simulations indicate that the oyster population can expand more rapidly than Dermo can expand and intensify, when the number of recruits is high (Fig. 7). Alternatively, or perhaps as an abetting process, the number of infective elements in the water column may be reduced below the level needed to generate an infective dose because of the volume of water filtered by the population at high abundance. An infective dose is hypothesized for MSX (Ford et al., 1999; Powell et al., 1999), and some evidence supports dose-dependency in Dermo (Bushek et al., 1997). However, insufficient information on the interaction of disease with oyster populations at high abundance is available to definitively decipher the rela- tionship between parasite and host at high abundance because oyster populations at high abundance are now rare or nonexistent for study. Interpretation and application of the compensatory and depensatory portions of the mortality curve de- scribed by Equation 6 (Fig. 9) come with a number of important caveats. 1) The probability of occurrence of an epizootic has increased since 1990 with the replacement of MSX by Dermo as the primary disease that produces mortality. An increase in frequency may be expected because of the greater tolerance of the parasite for low salinity (e.g., Ford, 1985; Powell et al., 1996; Ford et al., 1999; Ragone Calvo et al., 2001). Thus the ambit of oyster population dynamics may be more restricted by Dermo than by MSX. 2) The time series contains no high-abundance years since the replacement of MSX by Dermo circa 1990. Whether a return to high abundance is precluded by Dermo is unknown, but the difference in transmission dynamics between the two parasites (e.g., Ford and Tripp, 1996) and the expanded environ- mental range of Dermo in comparison to MSX would indicate that this may be the case. 3) Environmental conditions have not been constant over the 54 years, and environmental change significantly influences the chief agents of increased mortality, MSX and Dermo, as well as the autocorrelational dynamics of the epizootic process (e.g., Soniat et al., 1998). The mortality curve integrates environmental and biological dynamics. 4) The rise in winter temperature since the 1970s, that accelerated after 1990 (Scavia et al., 2002; Nixon et al., 2004), may have modified the interaction of disease with oyster population dynamics (e.g., Ford, 1996; Cook et al.; 1998, see also Hofmann et al., 1995; Powell et al., 1996), decreasing the applicability of the pre-1990 portion of the time series. 5) As abundance declines, a greater proportion of the oyster population is found on Table 6 One-year transition probabilities, as well as the fre- quency of occurrences of the eastern oyster ( Crassostrea virginica) population in each quadrant over the 54-yr time series were calculated from the Delaware Bay oyster recruitment-mortality distribution (Fig. 12). Median recruitment was 1.53x10® and the median mortality fraction was 0.127. Arrows indicate trajectories between quadrants. Quadrants are defined in Figure 10. Quadrant 1 2 3 4 1 — > 0.308 0.385 0.077 0.231 2 0.231 0.385 0.077 0.308 3 — > 0.143 0.071 0.714 0.071 4 0.231 0.231 0.154 0.385 Frequency of occurrence 0.241 0.259 0.259 0.241 Number of years 13 14 14 13 the medium-mortality beds (Powell et al., 2008). As a consequence, the probability of an epizootic begins to decline at abundances somewhere above 1 x 109 animals. Insufficient data are available to determine the trajec- tory for extrapolating this curve to lower abundances; therefore considerable uncertainty exists regarding the implementation of the abundance-mortality curve for abundances below 0.8 xlO9. Mortality and recruitment Both MSX and Dermo reduce the energy budget of a host (e.g., Hofmann et al., 1995; Ford et al., 1999) and, as a consequence, may reduce fecundity. Some empirical evidence exists that disease reduces the fecundity of individual oysters (Mackin, 1953; Barber et al., 1988; Ford and Figueras, 1988; Barber, 1996; Paynter, 1996; Dittman et al., 2001). One expectation is that fecundity may drop during epizootic years. No overall pattern is found between recruitment and box-count mortality in Delaware Bay (Ford and Figueras, 1988); however, the four massive settlement events with spat numbers above 1.5 xlO10 occurred during years when box-count mortality was very low, quadrant 3, (Fig. 12). Whether this coincidence is an independent outcome of two processes responding to common environmental and population forces, or whether it documents a causative connection, cannot yet be determined. Data points in the recruitment and box-count mor- tality distribution fell into quadrants 1-4 with a frequency of 13, 14, 14, and 13 years, respectively (Table 6). Such a distribution is expected by chance. Cases of high recruitment occur equally often with low and high mortality. Despite the seeming randomness of the relationship, mean first-passage times are far from equivalent across all transitions (Table 7). The high- recruitment+low-mortality state is reached from the other three quadrants about three times less frequently Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virginica in Delaware Bay 123 Figure 12 The relationship between box-count mortality and recruitment for the eastern oyster ( Crassostrea virginica ) in 1953-2006 for the natural oyster beds of Delaware Bay. Dotted lines indicate the 54-yr medians of box-count mortality and recruitment. than is any other population state. Once there, the population is much more like- ly to remain there than move to any of the other three quadrants. High recruit- ment with low mortality is a relatively stable state. Unrecorded mortality Box-count mor- tality is generally a measure of mortality of larger animals. Presumably, much of the mortality unrecorded by box counts is associated with predation in the first year of life and, therefore, likely would not show a discernible relationship with recruitment. Estimates of survival to one year of age indicate that mortality rates are at least a factor of three to five above the population average for older animals (Powell et al., 2009), confirming that much of the unrecorded mortality is juvenile mortality. The assumption that juvenile mortality rate varies randomly with respect to other indices of popula- tion dynamics is supported by compari- sons with abundance, recruitment, and box-count mortality (Figs. 13-15). Influence of regime shifts on biological relationships Both the broodstock abundance-recruit- ment (Fig. 7) and abundance-mortality (Fig. 9) curves have outlying points. These are more common in the former than in the latter. Arguably, data for years when regime shifts occur should not be used in defining such relationships because the purpose of such relationships is to understand and model the typi- cal population dynamics of the stock. Stock dynamics during regime shifts are atypical. The abundance-mortality relationship (Fig. 9A) shows only a single outlying point This outlier (X), the only case of epizootic mortality at stock abundances great- er than 5xl09, marks the regime shift year of 1985, when stock abundance reverted to the low-abundance state after more than a decade of high abundance. The 1968-70 period, during which time conditions supported a dramatic population expansion, did not leave an indel- ible imprint. All three years were characterized by low mortality, but many other such years displayed similar abundance levels. In contrast, the abundance-recruitment scatterplot (Fig. 7) contains four clear high-recruitment outliers. In this case, the 1985 regime shift is not unusual. Other low-recruitment years show high abundance. The 1968-70 period contains one of the four outliers (Fig. 7) and the years 1972-74 contain the other three. The inference drawn from Figure 2 is that these four outliers are of two types. One outlier is the previously- mentioned outlier that occurred during the 1968-70 period and represents the unusual event that dramati- Table 7 Mean first passage times and the distribution of occurrence of the eastern oyster ( Crassostrea virginica ) population in each quadrant after an infinite number of steps were calcu- lated from the Delaware Bay oyster recruitment-mortality distribution (Fig. 12). The observed distribution of occur- rences is given in Table 6.))Arrows indicate trajectories between quadrants. Quadrants are defined in Figure 10. Quadrant 1 2 3 4 Mean first passage time (yr) 4.45 3.72 9.97 4.52 2 -a- 4.83 3.79 9.91 4.17 3 — > 5.94 6.52 3.78 6.80 4 — > 4.92 4.65 9.08 4.05 Distribution of occurrence after infinite steps 0.225 0.264 0.264 0.247 cally impacted the stock. The other three are associated with an unusual transit of abundance above carrying capacity (Powell et al., 2008, 2009) and represent events that had no long-term consequences for the stock, ex- cept to maintain abundance near the carrying capacity 124 Fishery Bulletin 107(2) 0.0x10° 5.0x109 I.OxlO10 1.5x1010 2.0x1010 2.5x1010 Oyster (>20 mm) abundance 3.0x1010 3.5x1010 4.0x10'° Figure t3 The relationship between oyster abundance and unrecorded mortality for the eastern oyster ( Crassostrea virginica ) in 1953-2006 for the natural oyster beds of Delaware Bay. Irrational (positive) values on the ordinate indicate survey imprecision. -0.60 -0.40 -0.20 0.00 0.20 0.40 Fraction of population dying (unrecorded mortality) Figure 14 The relationship between box-count mortality and unrecorded mortality of the eastern oyster (Crassostrea virginica ), 1953-2006, for the natural oyster beds of Delaware Bay. Irrational (positive) values on the abscissa indicate survey imprecision. originally established circa 1970. In this scenario, these years were not unique. Nevertheless, for both cases, the performance of the stock was not representative of the dynamics defined by the remaining 50 years of observa- tion. As a consequence, a math- ematical relationship weighting these four observations overly much (e.g., the polynomial fit in Fig. 7) would not appropriately parameterize a model of the stock either in its high-abundance or low-abundance state. The influence of spatial relationships on population dynamics The relationships between brood- stock, recruitment, and mortality expressed by Equations 4 and 6 and by Figures 7 and 9 attempt to portray the time series of obser- vations in terms of the ambit of the stock’s population dynamics. In fact, in one sense, this mis- represents the true range of the species’ population dynamics at any particular time because the ambit of the stock in one regime differs from that of the other. First-passage times support this conclusion, as does a closer look at the distribution of abundance, recruitment, and mortality for the four bay regions over the full time series (Powell et al., 2008). Consider first the broodstock abundance-recruitment relation- ship (Fig. 7). We identify two sets of points characteristic of times when the stock was relatively consolidated within its distribu- tional range (Fig. 16). In these periods, a large proportion of the stock was found on the medium- mortality beds (Fig. 5). Such an occurrence was characteristic of the stock in both low- and high- abundance regimes and for ex- tended periods of time, including most years between 1960 and 1963 (low-abundance regime), the 1970-77 period (high-abundance regime), and the 1997-2006 pe- riod (low-abundance regime). The 1970s occurrences all fall in quadrant 4 (high abundance+high Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virginica in Delaware Bay 125 Q. CO O CD .Q E 5.0x1010- 4.5X1010- 4.0X1010- 3.5X1010- 3.0x1010- 2.5x1010- 2.0x1010- 1.5x10'°' ,10. 1 0x10 5.0x1 09- -0.80 -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 Fraction of population dying Figure 15 The relationship between unrecorded mortality and recruitment in 1953-2006 for the natural oyster beds of Delaware Bay. Irrational (positive) values on the abscissa indicate survey imprecision. recruitment) (Fig. 16). Where- as other years also fall in this quadrant, when the stock is con- solidated at high abundance, the likelihood that recruitment will be above the median of all years is extraordinarily high. When the stock is dispersed, the like- lihood is not as great, but still, in most years, the population’s performance falls into quadrant 4. Thus, stock dispersion has little influence on the outcome of recruitment events during the high-abundance regime. In contrast, occurrences when relatively more of the stock was found on the medium-mortal- ity beds during the low-abun- dance regime fall dispropor- tionately into quadrant 1 (low abundance + low recruitment) (Fig. 16). Eight of fourteen oc- currences in these years fall into this quadrant, a value sig- nificantly greater than expect- ed by an even distribution of points among the four quadrants (P<0.005), and 10 of 14 display low recruitment (quadrants 1 and 3), a value significantly greater than expected by an even split (P-0.05). Thus, when the stock is con- solidated within its range, and in its low-abundance regime, a high-recruitment event is unlikely. During the late 1980s and early 1990s, the stock was at rela- tively low abundance, but more distributed among bed regions (Fig. 5). These years are more evenly distrib- uted among the four quadrants (Fig. 16). In particu- lar, three occur in quadrant 3, accounting for a high percentage of all such events, and five fall above the long-term median for recruitment. Thus, although a dispersed stock can result in low recruitment during the low-abundance regime, the chance of a high-re- cruitment event is much improved. One inference from these data is that high recruit- ment events are the result of spawning by oysters down- bay of the medium-mortality region, in waters of higher salinity. This inference is supported by the tendency for the high-mortality beds to recruit more consistently (Powell et al., 2008). The fact that quadrants 1 and 4 are primarily represented by years when a consolidated stock distribution was present indicates that spawning potential differs between the two regimes. Perhaps it is no coincidence that the 1970 stock expansion was preceded by a tendency for the stock to expand at low abundance, thereby increasing the probability of a high recruitment event at low abundance. And perhaps it is no surprise that the decrease in recruitment during the first years of the 2000s (Powell et al., 2008) was preceded by a consolidation of the stock beginning in 1996, which reduced the probability of a high-recruit- ment event. All of these observations would imply that a high recruitment is primarily driven by increased spawning potential on higher-salinity beds. Thus, the broodstock-recruitment relationship (Fig. 7) fails to emphasize a substantive impact from stock dispersion. The range in recruitment at a given abundance, blithe- ly inferred to represent stochastic variation about a mean, in actuality includes a large influence from stock distribution that cannot be readily represented by a simple mathematical relationship between observed recruitment and stock size. This dispersion imprint is a dominant contributor to the dynamics of a population at low abundance, but not at high abundance, when compensatory processes begin to become important, and helps explain why the 1970 regime shift was an unlikely event. The influence of geographic dispersion is also ob- served in the abundance-mortality diagram (Fig. 9). Not surprisingly, the high-abundance regime is asso- ciated with low mortality, regardless of the degree of consolidation of the stock (Fig. 17). This association conforms with the relationships observed for the brood- stock-recruitment relationship (Fig. 16). By contrast, the years characterized by consolidated and dispersed stock during the low-abundance regime are divergent, and again this divergence is similar to our conclusion drawn from the broodstock-recruitment relationship. The mortality rate should be lower when more oysters are found on the medium-mortality beds, and this 126 Fishery Bulletin 107(2) Figure 16 The broodstock-recruitment relationship of the eastern oyster ( Crassostrea virginica), 1953-2006, for the natural oyster beds of Delaware Bay, showing low-abundance consolidated years (1960-1963 and 1996-2006), low-abundance dispersed years (1987— 1995), high-abundance consolidated years (1971-1978), high-abundance dispersed years (1979-1984), and low-abundance dispersed years (1987-1995). Consolidated and dispersed refer to the proportional contribution of the medium-mortality beds to total stock abundance as defined by the median value in Figure 5. The solid line is the best-fitted Ricker curve (Eq. 4). The dashed line is a second-order polynomial fit (see Kraeuter et ah, 2005). The dotted lines represent the 54-yr medians, which define the four quadrants (Fig. 10). Graph (A) represents the entire data set and (B) represents years with recruitment <6xl09 and abundance <5.5xl09. is the case. Epizootic mortalities (>0.20 in Fig. 17) occurred in four of eight years when the stock was dispersed, but only in three of fourteen years when consolidated. The second proportion differs signifi- cantly from the first (P<0.025). Epizootics are an important mechanism leading to stock consolidation, and a consolidated stock is resistant to further epi- zootic challenge. If one considers the relationships of recruitment and mortality to broodstock abundance, the high- Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virgimca in Delaware Bay 127 Figure 17 The broodstock-mortality relationship for the eastern oyster ( Crassostrea virginica ), 1953-2006, time period for the natural oyster beds of Delaware Bay, showing low-abun- dance consolidated years (1960-1963 and 1996-2006), high-abundance consolidated years (1971-1978), high-abundance dispersed years (1979-1984), and low-abundance dispersed years (1987-1995). The solid line is the curve fitted from Equation 6. The dotted lines represent the 54-yr medians which define the four quadrants (Fig. 10). Consolidated and dispersed refer to the proportion of the population on medium-mortality beds (Fig. 5) above or below the 54-yr median, respectively. The upper graph presents the entire dataset and the lower graph focuses on years of abundance <5 x 109. abundance regime is noteworthy for low mortality and high recruitment, regardless of stock disper- sion. However, during low-abundance intervals, the consolidated stock is in a relatively stable state and characterized by low mortality and low recruitment. The dispersed state is moderately less stable, char- acterized by higher mortality and higher recruit- ment. The interesting coincidence of similar trends in mortality and recruitment in both instances is noteworthy. 128 Fishery Bulletin 107(2) Conclusions The oyster population in Delaware Bay exhibits popu- lation dynamics that are not normally described in commercial species. One reason is the presence of multiple distinct, dynamically stable states delim- ited by temporally rapid regime shifts. Such dynam- ics are becoming more widely appreciated in fished species as a whole; therefore these unique dynamics may be more apparent than real. Oyster popula- tions display four unusual biological relationships, however, that impute greater peculiarity to their population dynamics. First, it seems likely that the broodstock-recruitment relationship, at least at low abundance, is driven more by the provision of settle- ment sites for larvae by the adults than by fecundity. Second, the natural mortality rate is temporally unstable and bears a nonlinear relationship with abundance (Fig. 9). This nonlinearity is driven by MSX and Dermo, both acting similarly despite the multifarious differences in their life histories, and by the environmental gradient of the habitable areas, which provide habitats of refuge from disease during epizootics. Third, high abundance and low mortal- ity, though likely requiring favorable environmental conditions, also seem to be self-reinforcing, although the specific underpinning dynamics remain unclear. As a consequence, an increased probability of high mortality occurs over a relatively small range of total abundances. The mortality relationship exhibits both compensatory and depensatory components. Fourth, the geographic distribution of the stock is inter- twined with the variables of abundance, recruitment, and mortality, such that biological relationships are functions both of spatial organization and inherent population processes. As a consequence of the imprint of geographic distribution on population dynamics, epizootic-level mortalities normally occur only when the animal has expanded its population beyond the refuge sufficiently that a significant fraction of the population is exposed to higher mortality. Consolida- tion limits mortality. What is equally interesting is the parallel influence on recruitment such that the consolidated stock has a lower recruitment potential, while also minimizing epizootic mortality. One is often dismayed by the dispersion of data in plots of the relationships of broodstock to recruitment and abundance to mortality. This dispersion is nor- mally ascribed to stochastic processes, and stochas- ticity is certainly a causal element. However, both governing regime and geographic distribution of the stock influence the dispersion of these data. Of note is the influence of stock dispersion, where the ambit of the population when the stock is in a contracted state is dissimilar from the ambit when the stock is in a dispersed state. This dynamic imposes a wider range in stock performance for a given stock abundance than would be observed for either distributional state alone. At least for oysters, a substantive component of ap- parent stochasticity observed in the relationships of recruitment and mortality to abundance originates not from simple year-to-year variation in stock perfor- mance, but from different distributions for the stock dictated by modifications in the geographic distribu- tion of the stock, and these distributional states tend to be self-reinforcing, as evidenced by similar changes in both recruitment and mortality over half-decadal or longer intervals of time. Acknowledgments We recognize the many people who contributed to the collection of survey data during the 54 years sur- veyed for this report, with particular recognition of H. Haskin, D. Kunkle, and B. Richards for their scientific contributions. We appreciate the many suggestions on content provided by S. Ford. The study was funded by an appropriation from the State of New Jersey to the Haskin Shellfish Research Laboratory, Rutgers University, and authorized by the Oyster Industry Sci- ence Steering Committee, a standing committee of the Delaware Bay Section of the Shell Fisheries Council of New Jersey. Literature cited Ackerman, E., L. R. Elveback, and J. P. Fox. 1984. Simulation of infectious disease epidemics, 202 p. Charles C. Thomas, Springfield, IL. Alexander, R. R., and G. P. Dietl. 2001. Shell repair frequencies in New Jersey bivalves: a recent baseline for tests of escalation with Tertiary, Mid-Atlantic congeners. Palaios 16:354-371. Allen, R. L. 1979. A yield model for the Foveaux Strait oyster ( Ostrea lutaria) fishery. Rapp. P.-V. Reun. Cons. Int. Explor. Mer 175:70-79. Anderson, R. M., and D. M. Gordon. 1982. Processes influencing the distribution of parasite numbers within host populations with special empha- sis on parasite-induced host mortalities. Parasitology 85:373-398. Andre, C., P. R. Jonsson, and M. Lindegarth. 1993. Predation on settling bivalve larvae by benthic suspension feeders: the role of hydrodynamics and larval behavior. Mar. Ecol. Prog. Ser. 97:183-192. Andreasen, V., and A. Pugliese. 1995. Pathogen coexistence induced by density-dependent host mortality. J. Theor. Biol. 177:159-165. Anonymous. 1996. Magnuson-Stevens Fishery Conservation and Management Act. U.S. Dep. Commer., NOAA Tech. Memo. NMFS-F/SPO-23, 121 p. Baines, P. G., and C. K. Folland. 2007. Evidence for a rapid global climate shift across the late 1960s. J. Climate 20:2721-2744. Barber, B. J. 1996. Gametogenesis of eastern oysters, Crassostrea virginica (Gmelin, 1791), and Pacific oysters, Crassostrea gigas (Thunberg, 1793) in disease-endemic lower Chesa- peake Bay. J. Shellfish Res. 15:285-290. Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virginica in Delaware Bay 129 Barber, B. J., S. E. Ford, and H. H. Haskin. 1988. Effects of the parasite MSX ( Haplosporidium nel- soni) on oyster ( Crassostrea virginica) energy metabolism. I. Condition index and relative fecundity. J. Shellfish Res. 7:25-31. Black, F. L. 1966. Measles endemicity in insular populations: critical community size and its evolutionary implications. J. Theor. Biol. 11:207-211. Botsford, L. W. 1981. The effects of increased individual growth rates on depressed population size. Am. Nat. 117:38-63. Breitburg, D. L., and R. S. Fulford. 2006. Oyster-sea nettle interdependence and altered con- trol within the Chesapeake Bay ecosystem. Estuaries Coasts 29:776-784. Burreson, E. M., N. A. Stokes, and C. S. Friedman. 2000. Increased virulence in an introduced pathogen: Haplosporidium nelsoni (MSX) in the eastern oyster Crassostrea virginica. J. Aquat. Anim. Health 12:1-8. Bushek, D., S. E. Ford, K. A. Alcox, R. Gustafson, and S. K. Allen Jr. 1997. Response of the eastern oyster, Crassostrea virgi- nica, to in vitro cultured Perkinsus marinus and the early fate of parasites delivered via three dosing methods. J. Shellfish Res. 16:479-485. Choi, J. S., K. T. Frank, W. C. Leggett, and K. Drinkwater. 2004. Transition to an alternate state in a continental shelf ecosystem. Can. J. Fish. Aquat. Sci. 61:505-510. Clark, W. G. 1999. Effects of an erroneous natural mortality rate on a simple age-structured stock assessment. Can. J. Fish. Aquat. Sci. 56:1721-1731. Collie, J. S., K. Richardson, and J. H. Steele. 2004. Regime shifts: can ecological theory illuminate the mechanisms? Prog. Oceanogr. 60:281-302. Cook, T., M. Folli, J. Klinck, S. Ford, and J. Miller. 1998. The relationship between increasing sea-surface temperature and the northward spread of Perkinsus marinus (Dermo) disease epizootics in oysters. Estuar. Coast. Shelf Sci. 46:587-597. Crocos, P. J. 1991. Reproductive dynamics of three species of Penaeidae in tropical Australia, and the role of reproductive studies in fisheries management. In Crustacean egg produc- tion (A. Wenner, and A. Kuris, eds.), p. 317-331. A. A Balkema, Rotterdam, Netherlands. Cummins, H., E. N. Powell, R. J. Stanton Jr., and G. Staff. 1986a. The rate of taphonomic loss in modern benthic habitats: how much of the potentially preservable com- munity is preserved? Palaeogeogr. Palaeoclimatol. Pal- aeoecol. 52:291-320. 1986b. The size-frequency distribution in palaeoecology: the effects of taphonomic processes during formation of death assemblages in Texas bays. Palaeontology (Lond.) 29:495-518. Dawe, E. G., E. B. Colbourne, and K. F. Drinkwater. 2000. Environmental effects on recruitment of short- finned squid illlex illecebrosus). ICES J. Mar. Sci. 57:1002-1013. Deakin, M. A. B. 1990. Catastrophe modelling in the biological sciences. Acta Biotheor. 38:3-22. DeAngelis, D. L., and J. C. Waterhouse. 1987. Equilibrium and nonequilibrium concepts in eco- logical models. Ecol. Monogr. 57:1-21. Dekshenieks, M. M., E. E. Hofmann, J. M. Klinck, and E. N. Powell. 2000. Quantifying the effects of environmental change on an oyster population: a modeling study. Estuaries 23:593-610. Dittman, D. E., S. E. Ford, and D. K. Padilla. 2001. Effects of Perkinsus marinus on reproduction and condition of the eastern oyster, Crassostrea virgin- ica, depend on timing. J. Shellfish Res. 20:1025— 1034. Fegley, S. R., S. E. Ford, J. N. Kraeuter, and H. H. Haskin. 2003. The persistence of New Jersey’s oyster seedbeds in the presence of MSX disease and harvest: management’s role. J. Shellfish Res. 22:451-464. Ford, S. 1997. History and present status of molluscan shellfish- eries from Barnegat Bay to Delaware Bay. In The his- tory, present condition, and future of the molluscan fisheries of North and Central America and Europe. Vol. 1, Atlantic and Gulf Coasts (C. L. MacKenzie Jr., V. G. Burrell Jr., A. Rosenfield, and W. L. Hobart, eds.). NOAA Tech. Rep. NMFS 127:119-140. Ford, S., E. Powell, J. Klinck, and E. Hofmann. 1999. Modeling the MSX parasite in eastern oyster ( Crassostrea virginica) populations. I. Model develop- ment, implementation, and verification. J. Shellfish Res. 18:475-500. Ford, S. E. 1985. Effects of salinity on survival of the MSX parasite Haplosporidium nelsoni (Haskin, Stauber, and Mackin) in oysters. J. Shellfish Res. 5:85-90. 1996. Range extension by the oyster parasite Perkinsus marinus into the northeastern United States: response to climate change. J. Shellfish Res. 15:45-56. Ford, S ,E., M. J. Cummings, and E. N. Powell. 2006. Estimating mortality in natural assemblages of oysters. Estuaries Coasts 29:361-374. Ford, S. E., and A. J. Figueras. 1988. Effects of sublethal infection by the parasite Hap- losporidium nelsoni (MSX) on gametogenesis, spawning, and sex ratios of oysters in Delaware Bay, USA. Diss. Aquat. Org. 4:121-133. Ford, S. E., and M. R. Tripp. 1996. Diseases and defense mechanisms. In The eastern oyster: Crassostrea virginica (V. S. Kennedy, R. I. E. Newell, and A. F. Eble, eds.), p. 581-659. Maryland Sea Grant College Program, College Park, MD. Frechette, M., and E. Bourget. 1985. Food-limited growth of Mytilus edulis L. in relation to the benthic boundary layer. Can. J. Fish. Aquat. Sci. 42:1166-1170. Frechette, M., and D. Lefaivre. 1990. Discriminating between food and space limita- tion in benthic suspension feeders using self-thinning relationships. Mar. Ecol. Prog. Ser. 65:15-23. Gabriel, W. L. 1992. Persistence of demersal fish assemblages between Cape Hatteras and Nova Scotia, northwest Atlantic. J. Northwest Atl. Fish. Sci. 14:29-46. Gill, C. A. 1928. The genesis of epidemics and the natural history of disease, 550 p. William Wood, New York, NY. Glover, C. P., and S. M. Kidwell. 1993. Influence of organic matrix on the post-mortem destruction of molluscan shells. J. Geol. 101:729— 747. 130 Fishery Bulletin 107(2) Godfray, H. C. J., and C. J. Briggs. 1995. The population dynamics of pathogens that control insect outbreaks. J. Theor. Biol. 176:125-136. Gray, J. S. 1977. The stability of benthic ecosystems. Helgol. Wiss. Meeresunters. 30:427-444. Hancock, D. A. 1973. The relationship between stock and recruitment in exploited invertebrates. Rapp. P.-V. Reun. Cons. Int. Explor. Mer 164:113-131. Haskin, H. H., and J. D. Andrews. 1988. Uncertainties and speculations about the life cycle of the eastern oyster pathogen Haplosporidium nelsoni (MSX). Am. Fish. Soc. Spec. Publ. 18:5-22. Haskin, H. H., and S. E. Ford. 1982. Haplosporidium nelson (MSX) on Delaware Bay seed oyster beds: a host-parasite relationship along a salinity gradient. J. Invertebr. Pathol. 40:388-405. Hastings, A. 1991. Structured models of metapopulation dynamics. In Metapopulation dynamics: empirical and theoretical investigations (M. Gilpin, and I. Hanski. eds.). Biol. J. Linn. Soc. 42:57-71. Heesterbeek, J. A. R, and H. G. Roberts. 1995. Mathematical models for microparasites of wildlife. In Ecology of infectious diseases in natural populations (B. T. Grenfell, and A. P. Dobson, eds.), p. 90-122. Cambridge Univ. Press, Cambridge, UK. Hilborn, R. 2002. The dark side of reference points. Bull. Mar. Sci. 70:403-408. Hilborn, R., and C. J. Walters. 1992. Quantitative fisheries stock assessment: choice, dynamics and uncertainty, 570 p. Chapman and Hall, New York, NY. Hoenig, J. M. 1983. Empirical use of longevity data to estimate mor- tality rates. Fish. Bull. 83:898-903. Hofmann, E. E., J. M. Klinck, J. N. Kraeuter, E. N. Powell, R. E. Grizzle, S. C. Buckner, and V. M. Bricelj. 2006. A population dynamics model of the hard clam, Mercenaria mercenaria : development of the age- and length-frequency structure of the population. J. Shell- fish Res. 25:417-444. Hofmann, E. E., E .N. Powell, E. A. Bochenek, and J. M. Klinck. 2004. A modelling study of the influence of environ- ment and food supply on survival of Crassostrea gigas larvae. ICES J. Mar. Sci. 61:596-616. Hofmann, E. E., E. N. Powell, J. M. Klinck, and G. Saunders. 1995. Modeling diseased oyster populations I. Modelling Perkinsus marinus infections in oysters. J. Shellfish Res. 14:121-151. Hofstetter, R. P. 1983. Oyster population trends in Galveston Bay 1973- 1978. Texas Parks and Wildl. Dept. Manag. Data Ser. 51, 33 p. Honkoop, P. J. C., J. van der Meer, J. J. Beukema, and D. Kwast. 1998. Does temperature-influenced egg production pre- dict the recruitment in the bivalve Macoma bal- thical Mar. Ecol. Prog. Ser. 164:229-235. Hughes, R. N. 1980. Optimal foraging theory in the marine context. Oceanogr. Mar. Biol. Annu. Rev. 18:423-481. Imeson, R. J., J. C. J. M. van den Bergh, and J. Hoekstra. 2002. Integrated models of fisheries management and policy. Environ. Model. Assess. 7:259-271. Jackson, J. B. C., M. X. Kirby, W. H. Berger, K. A. Bjorndal, L. W. Botsford, B. J. Bourque, R. H. Bradbury, R. Cooke, J. Erlandson, J. A. Estes, T. P. Hughes, S. Kidwell, C. B. Lange, H. S. Lenihan, J. M. Pandolfi, C. H. Peterson, R. S. Steneck, M. J. Tegner, and R. R. Warner. 2001. Historical overfishing and the recent collapse of coastal ecosystems. Science 293:629-637. Jaenike, J. 1998. On the capacity of macroparasites to control insect populations. Am. Nat. 151:84-96. Johnson, L. 1994. Pattern and process in ecological systems: a step in the development of a general ecological theory. Can. J. Fish. Aquat. Sci. 51:226-246. Kermack, W. O., and A. G. McKendrick. 1991. Contributions to the mathematical theory of epi- demics I. Bull. Math. Biol. 53:33-55. Knowlton, N. 2004. Multiple “stable states” and the conservation of marine ecosystems. Prog. Oceanogr. 60:387-396. Kraeuter, J. N., S. Buckner, and E. N. Powell. 2005. A note on a spawner— recruit relationship for a heavily exploited bivalve: the case of northern quahog (hard clams), Mercenaria mercenaria, in Great South Bay, New York. J. Shellfish Res. 24:1043-1052. Kraeuter, J. N., J. M Klinck, E. N. Powell, E. E. Hofmann, S. C. Buckner, R. E. Grizzle, and V. M. Bricelj. 2008. Effects of the fishery on the hard clam ( Mer- cenaria mercenaria ( L . )) population in Great South Bay, New York: a modeling study. J. Shellfish Res. 27:653-666. Link, J. S., J. K. T. Brodziak, S. F. Edwards, W. J. Overholtz, D. Mountain, J. W. Jossi, T. D. Smith, and M. J. Fogarty. 2002. Marine ecosystem assessment in a fisheries man- agement context. Can. J. Fish. Aquat. Sci. 59:1429- 1440. Lipcius, R. N., and W. T. Stockhausen. 2002. Concurrent decline of the spawning stock, recruit- ment, larval abundance, and size of the blue crab Cal- linectes sapidus in Chesapeake Bay. Mar. Ecol. Prog. Ser. 226:45-61. Livingston, R. J., F. G. Lewis, G. C. Woodsum, X.-F. Niu, B. Gal- perin, W. Huang, J. D. Christensen, M. E. Monaco, T. A. Bat- tista, C. J. Klein, R. L. Howard IV, and G. L. Ray. 2000. Modelling oyster population response to varia- tion in freshwater input. Estuar. Coast. Shelf Sci. 50:655-672. Loosanoff, V. L. 1966. Time and intensity of setting of the oyster Crassostrea virginica, in Long Island Sound. Biol. Bull. (Woods Hole) 130:211-227. MacKenzie Jr., C. L. 1996. History of oystering in the United States and Canada, featuring the eight greatest oyster estuaries. Mar. Fish. Rev. 58(4):l-78. Mackin, J. G. 1953. Incidence of infection of oysters by Dermocystidium in the Barataria Bay area of Louisiana. Natl. Shellfish. Assoc. Conv. Add. for 1951, p. 22-35. Mangel, M., B. Marinovic, C. Pomeroy, and D. Croll. 2002. Requiem for Ricker: unpacking msy. Bull. Mar. Sci. 70:763-781. Mangel, M., and C. Tier. 1994. Four facts every conservation biologist should know about persistence. Ecology 75:607-614. Powell et al.: Multiple stable reference points in oyster populations: Crassostrea virginica in Delaware Bay 131 Mann, R., and D. A. Evans. 1998. Estimation of oyster, Crassostrea virginica , stand- ing stock, larval production and advective loss in relation to observed recruitment in the James River, Virginia. J. Shellfish Res. 17:239-253. 2004. Site selection for oyster habitat rehabilitation in the Virginia portion of the Chesapeake Bay: a commentary. J. Shellfish Res. 23:41-49. Mann, R., J. S. Rainer, and R. Morales-Alamo. 1994. Reproductive activity of oysters, Crassostrea virgi- nica (Gmelin, 1791) in the James River, Virginia, during 1987-1988. J. Shellfish Res. 13:157-164. Maunder, M. N. 2003. Is it time to discard the Schaefer model from the stock assessment scientist’s toolbox. Fish. Res. 61:145-149. Maurer, D., and L. Watling. 1973. The biology of the oyster community and its associ- ated fauna in Delaware Bay. In Delaware Bay Report series vol. 6 (D. F. Polis, ed.), 97 p. College of Marine Studies, Univ. Delaware, Lewes, DE. Maurer, D., L. Watling, and R. Keck. 1971. The Delaware oyster industry: a reality? Trans. Am. Fish. Soc. 100:100-111. May, R. M., J. R. Beddington, J. W. Horwood, and J. G. Shepherd. 1978. Exploiting natural populations in an uncertain world. Math. Biosci. 42:219-252. McCann, K. S., L. W. Botsford, and A. Hasting. 2003. Differential response of marine populations to cli- mate forcing. Can. J. Fish. Aquat. Sci. 60:971-985. McGarvey, R., F. M. Serchuk, and I. A. McLaren. 1993. Spatial and parent-age analysis of stock-recruit- ment in the Georges Bank sea scallop ( Placopecten magellanicus ) population. Can. J. Fish. Aquat. Sci. 50:564-574. Milke, L. M., and V. S. Kennedy. 2001. Mud crabs (Xanthidae) in Chesapeake Bay: claw characteristics and predation on epifaunal bivalves. In- vertebr. Biol. 120:67-77. Moore, H. F. 1911. Condition and extent of the natural oyster beds of Delaware. U.S. Bur. Fish. 745:1-29. Myers, R. A., and N. J. Barrowman. 1996. Is fish recruitment related to spawner abun- dance? Fish. Bull. 94:707-724. Nixon, S. W., S. Granger, B. A. Buckley, M. Lamont, and B. Rowell. 2004. A one hundred and seventeen year coastal water temperature record from Woods Hole, Massachusetts. Estuaries 27:397-404. Osman, R. W., R. B. Whitlatch, and R. N. Zajac. 1989. Effects of resident species on recruitment into a community: larval settlement versus post-settlement mortality in the oyster Crassostrea virginica. Mar. Ecol. Prog. Ser. 54:61-73. Oviatt, C. A. 2004. The changing ecology of temperate coastal waters during a warming trend. Estuaries 27:895-904. Paloheimo, J. E. 1980. Estimation of mortality rates in fish popula- tions. Trans. Am. Fish. Soc. 109:378-386. Paynter, K. T. 1996. The effects of Perkinsus marinus infection on physi- ological processes in the eastern oyster, Crassostrea virginica. J. Shellfish Res. 15:119-125. Peterman, R. M. 1977. A simple mechanism that causes collapsing stabil- ity regions in exploited salmonid populations. J. Fish. Res. Board Can. 34:1130-1142. Peterson, C. H. 1984. Does a rigorous criterion for environmental identity preclude the existence of multiple stable points? Am. Nat. 124:127-133. Peterson, C. H., and R. Black. 1987. Resource depletion by active suspension feed- ers on tidal flats: influence of local density and tidal elevation. Limnol. Oceanogr. 32:145-166. Peterson, C. H., F. J. Fodrie, H. C. Summerson, and S. P. Powers. 2001. Site-specific and density-dependent extinction of prey by schooling rays: generation of a population sink in top-quality habitat for bay scallops. Oecologia ( Berl.) 129:349-356. Peterson, C. H., and H. C. Summerson. 1992. Basin-scale coherence of population dynamics of an exploited marine invertebrate, the bay scallop: implica- tions of recruitment limitation. Mar. Ecol. Prog. Ser. 90:257-272. Pierce, G. J., and J. G. Ollason. 1987. Eight reasons why optimal foraging theory is a complete waste of time. Oikos 49:111-117. Powell, E. N., K. A. Ashton-Alcox, S. E. Banta, and A. J. Bonner. 2001. Impact of repeated dredging on a Delaware Bay oyster reef. J. Shellfish Res. 20:961-975. Powell, E. N., K. A. Ashton-Alcox, J. A. Dobarro, M. Cummings, and S. E. Banta. 2002a. The inherent efficiency of oyster dredges in survey mode. J. Shellfish Res. 21:691-695. Powell, E. N., K. A. Ashton-Alcox, and J. N. Kraeuter. 2007. Reevaluation of eastern oyster dredge efficiency in survey mode: application in stock assessment. N. Am. J. Fish. Manag. 27:492-511. Powell, E. N., K. A. Ashton-Alcox, J. N. Kraeuter, S. E. Ford, and D. Bushek. 2008. Long-term trends in oyster population dynamics in Delaware Bay: regime shifts and response to disease. J. Shellfish Res. 27:729-755. Powell, E. N., E. A. Bochenek, J. M. Klinck, and E. E. Hofmann. 2002b. Influence of food quality and quantity on the growth and development of Crassostrea gigas larvae: a modeling approach. Aquaculture 210:89-117. 2004. Influence of short-term variations in food on sur- vival of Crassostrea gigas larvae: a modeling study. J. Mar. Res. 62:117-152. Powell, E. N., J. J. Gendek, and K. A. Ashton-Alcox. 2005. Fisherman choice and incidental catch: size fre- quency of oyster landings in the New Jersey oyster fishery. J. Shellfish Res. 24:469-476. Powell, Eric N., John M. Klinck, K. A. Ashton-Alcox, and John N. Kraeuter. 2009. Multiple stable reference points in oyster populations: implications for reference point-based management. Fish. Bull. 107:133-147. Powell, E. N., J. M. Klinck, S. E. Ford, E. E. Hofmann, and S. J. Jordan. 1999. Modeling the MSX parasite in Eastern oyster ( Crassostrea virginica) populations. III. Regional appli- cation and the problem of transmission. J. Shellfish Res. 18:517-537. Powell, E. N., J. M. Klinck, and E. E. Hofmann. 1996. Modeling diseased oyster populations. II. Trigger- 132 Fishery Bulletin 107(2) ing mechanisms for Perkinsus marinus epizootics. J. Shellfish Res. 15:141-165. Powell, E. N., J. M. Klinck, E. E. Hofmann, and M. A. McManus. 2003. Influence of water allocation and freshwater inflow on oyster production: a hydrodynamic-oyster (sic) popu- lation model for Galveston Bay, Texas, USA. Environ. Manag. 31:100-121. Powell, E. N., J. M. Klinck, E. E. Hofmann, and S. M. Ray. 1994. Modeling oyster populations. IV: rates of mortal- ity, population crashes and management. Fish. Bull. 92:347-373. Powell, E. N., J. M. Klinck, E. E. Hofmann, E. A. Wilson-Ormond, and M. S. Ellis. 1995. Modeling oyster populations. V. Declining phyto- plankton stocks and the population dynamics of Ameri- can oyster ( Crassostrea uirginica) populations. Fish. Res. 24:199-222. Powell, E. N., R. J. Stanton Jr., D. Davies, and A. Logan. 1986. Effect of a large larval settlement and catastrophic mortality on the ecologic record of the community in the death assemblage. Estuar. Coast. Shelf Sci. 23:513-525. Pyke, G. H. 1984. Optimal foraging theory: a critical review. Annu. Rev. Ecol. Syst. 15:523-575. Ragone Calvo, L. M., R. L. Wetzel, and E. M. Burreson. 2001. Development and verification of a model for the population dynamics of the protistan parasite, Per- kinsus marinus, within its host, the Eastern oyster, Crassostrea virginica, in Chesapeake Bay. J. Shellfish Res. 20:231-241. Redner, S. 2001. A guide to first-passage processes, 312 p. Cam- bridge Univ. Press, Cambridge, U.K. Restrepo, V. R., G. G. Thompson, P. M. Mace, W. K. Gabriel, L. L. Low, A. D. MacCall, R. D.Methot, J. E. Powers, B. L. Taylor, P. R. Wade, and J. F. Witzig. 1998. Technical guidance on the use of precautionary approaches to implementing National Standard 1 of the Magnuson/Stevens Fishery Conservation and Man- agement Act. NOAA Tech. Memo. NMFS-F/SPO-31, 54 p. Rheault, R. B., and M. A. Rice. 1996. Food-limited growth and condition index in the Eastern oyster, Crassostrea virginica. (Gmelin 1791), and the bay scallop, Argopecten irradians irra- dians (Lamarck 1819). J. Shellfish Res. 15:271— 283. Rice, J. 2001. Implications of variability on many time scales for scientific advice on sustainable management of living marine resources. Prog. Oceanogr. 49:189-209. Rothschild, B. J. 2000. “Fish stocks and recruitment”: the past thirty years. ICES J. Mar. Sci. 57:191—201. Rothschild, B. J., C. Chen, and R. G. Lough. 2005. Managing fish stocks under climate uncertainty. ICES J. Mar. Sci. 62:1531-1541. Rothschild, B. J., and A. J. Mullen. 1985. The information content of stock-and-recruitment data and non-parametric classification. J. Cons. Cons. Int. Explor. Mer 42:116-124. Rothschild, B. J., and L. J. Shannon. 2004. Regime shifts and fishery management. Prog. Oceanogr. 60:397-402. Sakuramoto, K. 2005. Does the Ricker or Beverton and Holt type of stock-recruitment relationship truly exist? Fish. Sci. 71:577-592. Scavia, D., J. C. Field, D. F. Boesch, R. W. Buddemeier, V. Burkett, D. R. Cayan, M. Fogarty, M. A. Harwell, R. W. Howarth, C. Mason, D. J. Reed, T. C. Royer, A. H. Sallenger, and J. G. Titus. 2002. Climate change impacts on U.S. coastal and marine ecosystems. Estuaries 25:149-164. Sissenwine, M. P. 1984. Why do fish populations vary? In Exploita- tion of marine communities (R. M. May, ed.), p. 59— 94. Springer-Verlag, Berlin. Sissenwine, M. P., and J. G. Shepherd. 1987. An alternative perspective on recruitment over- fishing and biological reference points. Can. J. Fish. Aquat. Sci. 44:913-918. Soniat, T. M., E. N. Powell, E. E. Hofmann, and J. M. Klinck. 1998. LTnderstanding the success and failure of oyster populations: the importance of sampled variables and sample timing. J. Shellfish Res. 17:1149-1165. Soniat, T. M., E. N. Powell, J. M. Klinck, and E. E. Hofmann. In press. LTnderstanding the success and failure of oyster populations: climatic cycles and Perkinsus marinus. Int. J. Earth Sci. Southworth, M., and R. Mann. 1998. Oyster reef broodstock enhancement in the Great Wicomico River, Virginia. J. Shellfish Res. 17:1101-1114. 2004. Decadel scale changes in seasonal patterns of oyster recruitment in the Virginia sub estuaries of the Chesapeake Bay. J. Shellfish Res. 23:391-402. Steele, J. H., and E. W. Henderson. 1984. Modeling long-term fluctuations in fish stocks. Sci- ence. 224:985-987. Tamburri, M. N., R. K. Zimmer, and C. A. Zimmer. 2007. Mechanisms reconciling gregarious larval settle- ment with adult cannibalism. Ecol. Monogr. 77:255-268. Vetter, E. F. 1987. Estimation of natural mortality in fish stocks: a review. Fish. Bull. 86:25-43. Wallace, R. K., W. Hosking, and S. T. Szedlmayer. 1994. Fisheries management for fishermen: a manual for helping fishermen understand the federal management process. NOAA MASG P-94-012, 56 p. Ware, D. M. 2000. Aquatic ecosystems: properties and models. In Fisheries oceanography: an integrative approach to fisheries ecology and management (P. J. Harrison, and T. R. Parsons, eds.), p. 163-194. Blackwell Science, Oxford, England. Whitlatch, R. B., and R. W. Osman. 1994. A qualitative approach to manage shellfish populations: assessing the relative importance of tro- phic relationships between species. J. Shellfish Res. 13:229-242. Wilson-Ormond, E. A., E. N. Powell, and S. M. Ray. 1997. Short-term and small-scale variation in food avail- ability to natural oyster populations: food, flow and flux. Mar. Ecol. 18:1-34. 133 Multiple stable reference points in oyster populations: implications for reference point-based management Eric IN. Powell (contact author)1 John JVS. KSinck2 Kathryn A. Ashton-Alcox1 John N. Kraeuter1 Email address for contact author: eric@hsrl.rutgers.edu 1 Haskin Shellfish Research Laboratory Rutgers University 6959 Miller Ave. Port Norris, New Jersey 08349 2 Center for Coastal Physical Oceanography Crittenton Hall Old Dominion University Norfolk, Virginia 23529 Abstract — In the second of two com- panion articles, a 54-year time series for the oyster population in the New Jersey waters of Delaware Bay is analyzed to examine how the pres- ence of multiple stable states affects reference-point-based management. Multiple stable states are described by four types of reference points. Type I is the carrying capacity for the stable state: each has associ- ated with it a type-II reference point wherein surplus production reaches a local maximum. Type-II reference points are separated by an intermedi- ate surplus production low (type III). Two stable states establish a type-IV reference point, a point-of-no-return that impedes recovery to the higher stable state. The type-II to type-III differential in surplus production is a measure of the difficulty of rebuild- ing the population and the sensitivity of the population to collapse at high abundance. Surplus production pro- jections show that the abundances defining the four types of reference points are relatively stable over a wide range of uncertainties in recruitment and mortality rates. The surplus pro- duction values associated with type- II and type-III reference points are much more uncertain. Thus, biomass goals are more easily established than fishing mortality rates for oyster populations. Manuscript submitted 29 November 2007. Manuscript accepted 9 September 2008. Fish. Bull. 107:133-147 (2009). The views and opinions expressed or implied in this article are those of the author and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. All federal fisheries, and some state fisheries, are managed under biologi- cal reference-point guidelines under which a specific yearly allocation or quota is advised to constrain fishing mortality (e.g., Wallace et al.1). The biological reference-point approach for federal fisheries mandated by the Magnuson-Stevens Fishery Conser- vation and Management Act (Anony- mous, 1996) requires management of fish populations at a biomass that provides maximum sustainable yield. In this system, sophisticated survey, analytical, and modeling procedures are used to identify selected biological reference points, such as the target biomass, Bmsy, and the carrying capac- ity, K. Fishing mortality is then set in relation to reference point goals. Normally, B is defined in relation to carrying capacity, the biomass present without fishing, where natu- ral mortality balances recruitment (e.g., May et al., 1978; Johnson, 1994; Mangel and Tier, 1994; Rice, 2001). This stable point is characterized by a population for which most animals are adults, where natural mortality rates are low, and where recruitment is limited by compensatory processes such as resource limitation constrain- ing fecundity. Bm is most commonly defined as based on the well-known Schaefer model that stipulates the guiding premise that surplus pro- duction is highest at y (Hilborn and Walters [1992]; see Restrepo et al. [1998] for more details on the federal management system). The raison d’etre for reference- point-based management is the de- velopment of equilibria between re- cruitment (and growth) and mortality at target host densities (the archetype being Bmsy). Unfortunately, for man- aging oyster populations, obstacles ex- ist in meeting this objective because oyster populations do not appear to be inherently equilibrious, particularly those subjected to MSX, a disease caused by the protozoan Haplospo- ridium nelsoni, or Dermo, a disease caused by the protozoan Perkinsus marinus. Time series of oyster abun- dance typically show wide interan- nual variations, mediated in no small measure by year-to-year differences in natural mortality rate, although overfishing has also been an impor- tant contributing agent (e.g., Mann et al., 1991; Rothschild et ah, 1994; Burreson and Ragone Calvo, 1996; Ragone Calvo et al., 2001; Jordan et 1 Wallace, R. K., W. Hosking, and S. T. Szedlmayer. 1994. Fisheries manage- ment for fishermen: A manual for help- ing fishermen understand the federal management process. NOAA MASG P-94-012, 56 p. 134 Fishery Bulletin 107(2) al., 2002; Powell et al., 2008). In the first of two com- panion contributions, we described the case for oyster populations in Delaware Bay. A 54-year time series documents two regime shifts, circa-1970 and circa-1985, with intervening and succeeding intervals having the attributes of alternate stable states ( sensu Gray, 1977; Peterson, 1984; Knowlton, 2004). Within these periods are substantial population excursions produced by vary- ing rates of recruitment and natural mortality, but the alternate stable states are demarcated by even larger excursions in abundance. Moreover, these periods of relative stability delineated by regime shifts are per- sistent and transcend a range of climatic conditions (Soniat et al., in press). Population dynamics of the Delaware Bay oyster pop- ulation is not solely a function of disease, but stable- point abundances are at least partially a byproduct of disease, and disease has played a role in regime shifts (Powell et al., 2008). The classic view of carrying ca- pacity fails when disease accounts for a substantial fraction of natural mortality and this compromises an estimate of Bm Some have attempted to redefine car- rying capacity in diseased populations in relation to the abundance (population density in classic disease models, e.g., Kermack and McKendrick [1991], Hethcote and van den Driessche [1995]) at which each diseased animal will produce, in its lifetime, a single infection event (e.g., Heesterbeek and Roberts, 1995; Swinton and Anderson, 1995). This abundance is always below, usually well below, the original K. When abundance rises above this level, the influence of disease increases, as does the chance of epizootic mortality. This increase restrains population abundance below the predisease K (e.g., Kermack and McKendrick, 1991; Hasiboder et al., 1992; Godfray and Briggs, 1995; Frank, 1996). This approach is not well tailored to diseases such as MSX and Dermo for which environment is a potent modu- lator of effect and in which rapid transmission rates are not requiring of host-to-host contact. Furthermore, the existence of multiple apparently stable states and regime shifts imply that the standard Schaefer model, from which such basic biological references points as B are derived, also does not provide the appropriate framework for managing oyster populations because this model has only a single stable state. These ratiocinations lead to three salient questions pertinent to developing management goals for oyster stocks: 1) Can reference points be defined that consis- tently permit fishing without jeopardizing the sustain- ability of the stock? 2) Must management goals be set within the context of each of several multiple stable states? 3) How does regime change affect the usefulness of reference points and can management goals be set to increase the probability of regime shift to a preferred stable state? In this contribution, we use the case of the Delaware Bay oyster stock in New Jersey waters to examine these questions. In a companion contribution, we describe the long-term survey time series and the relationships of broodstock abundance with recruitment and mortality (Powell et al., 2009). In this contribu- Table 1 The bed groups (by location: upbay and downbay) and subgroups (by mortality rate) for the eastern oyster ( Crassostrea virginica) collected on twenty beds in Delaware Bay, as shown in Figure 1. Mortality rate divides each of the primary groups, themselves being divided by location, a surrogate for upbay-downbay vari- ations in dredge efficiency and fishery-area management regulations. Bed group/subgroup Bed name Upbay group Low mortality Medium mortality Downbay group Medium mortality High mortality Round Island, Upper Arnolds, Arnolds Upper Middle, Middle, Sea Breeze, Cohansey, Ship John Shell Rock Bennies Sand, Bennies, New Beds, Hog Shoal, Hawk’s Nest, Strawberry, Vexton, Ledge, Egg Island, Nantuxent Point, Beadons tion, we develop a surplus production model to relate these relationships with stock performance over a range of abundances. Following discussion of the results of simulations with this model, we consider the basis for an MSY- based management system for oyster popula- tions and the implications of multiple stable states in the decision-making process. Model formulations and statistics Powell et al. (2008, 2009) have provided an overview of the oyster populations in Delaware Bay during the 1953-2006 time period. Analyses of the Delaware Bay oyster resource of New Jersey routinely reveal a divi- sion between the upbay group of eight beds (Round Island, Upper Arnolds, Arnolds, Upper Middle, Middle, Sea Breeze, Cohansey, and Ship John [Fig. 1]) and the downbay group of twelve beds (Shell Rock, Bennies Sand, Bennies, New Beds, Nantuxent Point, Hog Shoal, Hawk’s Nest, Strawberry, Vexton, Beadons, Egg Island, and Ledge). Salinity, natural mortality rate, and growth rate are higher downbay. Dredge efficiencies are signifi- cantly higher downbay (Powell et al., 2002, 2007). Both regions can be subdivided on the basis of natural mortal- ity rate and productivity. In the upbay group, natural mortality rates and growth rates are significantly lower for the upper three beds, Round Island, Upper Arnolds and Arnolds, than for the remaining beds. Henceforth these two groups will be termed the low-mortality and medium-mortality beds (Table 1). In the downbay group, growth rates and mortality rates are lower for Shell Rock, leading to its designation as a medium-mortality bed; the reminder are high-mortality beds (Table 1). Powell et al.: Multiple stable reference points in oyster populations 135 75° 30' 75° 20' 75° 10' 39° 25' 39° 20 39° 39° 10' 39“ 25’ 39° 20' 39° 15' 39° 10' 75° 20' 75° 10' Figure 1 The twenty natural oyster beds of the eastern oyster (Crassostrea virginica) in the New Jersey waters of Delaware Bay may be characterized in terms of high-quality (dark shade) and medium-quality (light shade) grids. The term “quality” refers to a relative differential in long-term average oyster abundance (Powell et al, 2008). The footprints for the Middle bed (upper portion of figure) and the beds downbay from it, exceptNew Beds, Egg Island, and Ledge, were updated with data from surveys in 2005 and 2006. The footprints for the remaining beds were based on historical definitions. Throughout this contribution, we will refer to these bay regions where necessary, but in general, we will model the entire stock. In the following section, we summarize the biological relationships identified by Powell et al. (2009) without further discussion. Natural mortality fractions were obtained from box counts (be) under the assumption that ^ oysterSf_i ~ ^ boxeSf + ^ live oysterst > HI where N = the number of individuals. Hence, = NboxeSt (2) bc N, + N, boxes f live oysters f where 0, is obtained by difference: O0 = ( Nt - Nt_, ) -(/?,_! - <&6cAU - ®fNt_ x) Nf~i + Rt-i (4) where 0, and ^4<0. We will refer to maxima in surplus production as type-II reference points (Fig. 3). Because the time series under analysis is configured in terms of abundance rather than biomass, the designation Nmsy, rather than Bmsy, will be used hereafter. We present hereafter a series of simulations of the Delaware Bay oyster stock designed to examine the change in surplus production with abundance. We first consider a population for which recruitment rate fol- lows Equation 9, a compensatory curve, with a 54-yr average unrecorded mortality rate (Eq. 5), and with the box-count mortality rate described by Equation 12. 138 Fishery Bulletin 107(2) These relationships are depicted in Figures 7 and 10 of Powell et al. (2009). The trajectory for surplus pro- duction under these constraints is compared in Figure 3 and detailed in Figure 4. Recruitment rate rises as abundance declines (Fig. 4). This is anticipated from the compensation inherent in the relationship between broodstock and recruitment. The box-count mortal- ity rate shows a maximum somewhat above an abun- dance of 2xl09 (Fig. 4). These relationships define a trend between surplus production and abundance that is divergent from the normal Schaefer curve (Ricker, 1975; Hilborn and Walters, 1992; Haddon, 2001; Zabel, 2003), as expected. The single type-I reference point is at iV=9.3xl09. This is an estimate of carrying capac- ity, K. Typically a single type-II reference point would exist, Nmsy, at about y, but in this case two maxima in surplus production exist, one higher, N^nsy, than the other, NLmsy. N^lsy is at iV=4.86 x 109. This is the abundance classically interpreted as Nmsy, and, indeed, surplus production is maximal at this point and the value is approximately -j- The second type-II reference point occurs at A=1.43 x 109. Unlike the simple Schaefer curve depicted in Hilborn and Walters (1992), Haddon (2001), and Zabel (2003), a local minimum in surplus production exists between these two type-II surplus production maxima, at 1V=2.57 x 109. In this case, sur- plus production remains above zero, St> 0. An increase in abundance above this level and a decrease in abun- dance below this level both increase surplus produc- tion. This reference point, herein designated type III, always occurs between two maxima in surplus produc- tion and is characterized by -s-=0 and 0 (Table 2). dN dN The unusual nature of the surplus production curve in Figure 4, that yields the local minimum in surplus production and a secondary surplus production peak at a lower abundance, is produced by the depensatory and compensatory segments of the box-count mortality relationship established by the relationship between the occurrence of epizootics and abundance in the Delaware Bay oyster stock. Figures 5-7 show three alternative trajectories for the change in surplus production with abundance in the Delaware Bay oyster stock obtained by small modi- fications of the parameters governing recruitment and mortality. The first is obtained by using the 54-yr me- dian unrecorded mortality rate, rather than the 54-yr mean rate. The median is distinctly higher. Again, the surplus production trajectory includes one type-I, two type-II, and one type-III reference points (Figs. 3 and 5). The abundance associated with the four refer- ence points remains unchanged, although the surplus production values associated with the type-II maxima and type-III minimum are lower than in the preceding case (Table 2). The second alternative is obtained after a perusal of Figure 10 in Powell et al. (2009) that shows that the mortality rate for stock abundances frequented by epi- zootics often falls below the curve provided by Equation 12. This is a function of stock dispersion that modulates the likelihood of epizootic mortality rates (Powell et al., 2009). In fact, on the average, box- count mortality rate reaches epizo- otic levels only half the time. Thus, Figures 3 and 6 show the trend in surplus production when epizoot- ics are assumed to occur only half the time, and box-count mortality rate is expressed as the average of a year with an epizootic and a year without one. The type-III reference point is nearer the N^sy value in this surplus production trajectory, so that the valley between NL and is something more than a shoulder on the surplus production curve. Thus, the value of the sur- plus production maxima, averaged over a number of years, is strongly influenced by the frequency and in- tensity of epizootics (Table 2). The final alternative addresses the uncertainty that exists in the shape of the broodstock-recruit- ment curve at low abundance. Linearizing the curve at low abun- dance (Eq. 11) yields a surplus production trajectory depicted in Figure 8 of Powell et al. (2009). The relationship is unique in gen- erating a second type-I reference Powell et al. : Multiple stable reference points in oyster populations 139 point, at V=1.93 x 109. This is a multiple-stable-point system with two carrying capacities, one at KH and one at KL. Note that the lower surplus production maximum is closer to KL than expected by the Schaefer relationship: NLmsy> ^ (Fig. 7). This representation of oyster population dynamics also gener- ates a type-IV reference point at A/=3.03xl09. Type IV, like type I, is characterized by S.-O and 0, but in this case -ttt >0 (Table 2). Figure 8 presents a stylized version of the surplus production trajectory of Figure 7. Note that the type-I reference points are points of con- vergence. Abundance rising above this value will produce negative surplus production and a return to the abundance level and vice ver- sa for a decline in abundance. On the other hand, type-IV reference points are divergences or points of population instability. They mark thresholds for population collapse. The divergence that is the type-IV reference point is maintained by the competing rates of box-count mortality and recruitment that switch in dominance at this point (Fig. 8). A population reaching a type-IV reference point as abun- dance declines will see a rapid fur- ther decline. Once below this point, the likelihood becomes very low that the population can cross the gulf and re-acquire its high-abundance trajectory. Reference-point-based management Carrying capacity Perusal of the time series suggests that population abundances above about 12xl09 are unstable. The analyses provided using Equation 14 return this same expectation, that carrying capacity is about 9.3 x 109. This explains the stability of population abundance during the 1970s as the population was at or near carrying capacity (Fig. 9). Abundance rose above this point a number of times between 1970 and 1985, but higher abundances were not sustainable. Interest- ingly, this carrying capacity is a carrying capacity for a population enzootic for MSX disease. The natural mortality rate during the 1970s is not much different from the few measures that exist for the time frame pre-1957 and the pre-MSX years are not outliers on the broodstock-recruitment diagram. So, MSX was not a significant agent of mortality during this period. Hence, predisease carrying capacity for which no empirical quantitative record exists is likely to have been similar to abundances during the 1970s, with the observed dif- 2 4 6 8 10 Abundance (billions) Figure 4 The relationship of surplus production (Eq. 14), the rates of recruitment, unrecorded mortality, box-count mortality, and a conditional estimate of catch expressed as the fraction of the stock, for parameters defined by, for recruit- ment, T( from Equation 10, m0 from Equation 5 using the 54-year average 4>0, and mbc from Equation 12. This simulation assumes compensation in the broodstock-recruitment curve, average unrecorded (mostly juvenile) mortal- ity, and a box-count mortality rate that emphasizes epizootic mortality at low abundance. Catch estimates are conditional on the assumption of long-term persistence of a chosen abundance level and distribution of the entire stock in habitats permitting growth to market size. Table 2 The surplus production values associated with the types I, II, III, and where applicable, IV reference points depicted in the referenced figures and the defining characteris- tics of each reference point type. Surplus production is expressed in billions of oysters. NA=not applicable. Figure number Type I Carrying capacity (. K ) Type II NH msy Type III c ^ min Type II NL msy Type IV Point of no return Surplus production 4 0.0 0.665 0.167 0.319 NA 5 0.0 0.511 0.103 0.275 NA 6 0.0 0.519 0.297 0.318 NA 7 0.0 0.511 -0.094 0.112 0.0 c dS d2S dN dN2 Defining characteristics Type I = 0 <0 ~0 Type II >0 = 0 <0 Type III >0 or <0 = 0 >0 Type IV = 0 >0 ~0 140 Fishery Bulletin 107(2) ferential in abundance in the 1950s primarily a result of the higher fishing mortality rates during that time. Carrying capacity is defined by a set of criteria that are normally thought to be unique (Table 2). Inter- estingly, in Delaware Bay oyster populations, a sec- ond type-I reference point may exist, depending on the presence of a reference point of type IV, as considered subsequently. This type-I reference point, if present, is at 1.93 xlO9, nearly a factor of 5 lower in abundance than the classic carrying capacity. However, this value is also similar to the abundance observed during the low-abundance phase of the population (Fig. 9), an out- come anticipated of a population with multiple stable points (Gray, 1977; Peterson, 1984) in which community compositions are theorized to resolve themselves into preferred states that can be exchanged only through triggering mechanisms capable of overcoming the iner- tia of the individual states. Soniat et al. (1998) argued that inertia is an important attribute of oyster popula- tion dynamics and that this inertia minimizes the in- fluence of short-term environmental shifts. The 54-year time series of Delaware Bay supports the importance of inertia and suggests some reasons for how population dynamics are internally stabilized. Both recruitment and mortality have abundance-de- pendent rates. The high-abundance regime is inherently stable. Mean first passage times ( sensu Rothschild and Mullen, 1985; Redner, 2001; Rothschild et al., 2005) for transitions to the alternate stable state typically exceed 6 yr (Powell et al., 2009). Given a population at high abundance: that population will tend to maintain itself because high abundance, on the average, gener- ates higher recruitment, and also, on the average, is associated with lower rates of natural mortality. Thus, high abundances have a strong internal self-sustain- ing mechanism. However, the 1970-85 period occurred prior to the onset of Dermo disease in Delaware Bay. Whether a high abundance state is sustainable under any environmental conditions with Dermo as the prin- cipal agent of mortality is unknown. The low-abundance regime is stable only if the sur- plus production minimum separating the two maxima is negative. The differential between the two carrying capacities, KH and KL, is a factor of 4.82. Powell et al. (2009) discuss the tendency for the Delaware Bay oyster population to contract to a habitat of refuge on the me- dium-mortality beds (Table 1) as abundance falls. This occurs due to the gradient in natural mortality that increasingly penalizes the popula- tion downestuary. The differential in bed area between the entire bay and the medium-mortality beds is a factor of 2.46 excluding the two lowermost and least produc- tive beds, Egg Island and Ledge, or 2.70 including them. Thus, habitat area, though likely a contributor to the differential in the two car- rying capacities, does not explain adequately the differential between KL and KH, and this agrees with the observation (Figure 5 in Powell et al., 2009) that contracted and dispersed population distributions both prevailed for extended periods during the low-abundance regime. Surplus production targets Bever- ton et al. (1984) distinguish between short-term catch forecasts used to generate a yearly TAL and long- term strategic assessments used to set abundance goals. The constant- abundance reference point imple- mented with the model of Klinck et al. (2001) is particularly useful in maintaining a population close to an abundance target and has been used for short-term catch forecasts but does not lend itself to long-term strategic assessments. The purpose of this study was to develop refer- ence points that might be used to set abundance goals. 0.6-i ,-0.6 0 2 4 6 8 10 12 Abundance (billions) Figure 5 The relationship of surplus production (Eq. 14), the rates of recruitment, unrecorded mortality, and box-count mortality, and a conditional estimate of catch expressed as the fraction of the stock, for parameters defined by, for recruitment, Tt from Equation 10, m0 from Equation 5 using the 54-year median 4>0, and mbc from Equation 12. This simulation assumes compensation in the broodstock-recruitment curve, median unrecorded (mostly juvenile) mortality, and a box-count mortality rate that emphasizes epizootic mortality at low abundance. This simulation differs from the simulation in Figure 4 in a higher level of unrecorded mortality. Catch estimates are conditional on the assumption of long-term persistence of a chosen abundance level and distribu- tion of the entire stock in habitats permitting growth to market size. Powell et al.: Multiple stable reference points in oyster populations 141 0.6-1 ,-0.6 Abundance (billions) Figure 6 The relationship of surplus production (Eq. 14), the rates of recruitment, unrecorded mortality, and box-count mortality, and a conditional estimate of catch expressed as the fraction of the stock, for parameters defined by, for recruitment, T, from Equation 10, m0 from Equation 5 using the 54-year median 0, and mhc from Equation 12. This simulation assumes a linear relationship between broodstock abundance and recruitment at low abundance, median unrecorded (mostly juvenile) mortality, and a box-count mortality rate that emphasizes epizootic mortality at low abundance. In comparison to simulations depicted in Figures 4-6, this simulation has a combination of relatively high natural mortality and relatively low recruitment. Catch estimates are conditional on the assumption of long-term persistence of a chosen abundance level and dis- tribution of the entire stock in habitats permitting growth to market size. Powell et al.: Multiple stable reference points in oyster populations 143 tion goals and methods have received consider- able attention (e.g., Breitburg et al., 2000; Mann, 2000). Restoration goals are dramatically impacted by the location of type-II reference points in relation to stock abundance. Type II is the goal under MSY management objectives, and by the presence of type IV and the differential between types II and III. The difference between type II and type III affects 1) the ease of transition from one sta- ble point to another and 2) the impact on fishery yield during the transition. As the differential increases, from the example in the surplus pro- duction trajectory of Figure 6 to that in Figure 7 for instance, the limitation on fishery yield during the transition must increase. The obvious incon- gruity will be an observed increase in abundance of marketable stock during times of decreased al- location necessitated by the transitory limitation on surplus production coincident with the type-III reference point. This apparent inequity will likely exacerbate the natural adversarial relationship that exists between regulator and industry. The frequently complex relationship between economics and biology in fisheries management is well known (Lipton and Strand, 1992; Mackinson et al., 1997; Imeson et al., 2002). Thus several questions come to the fore. Can rebuilding to N1^ be accomplished? This depends on the existence of type IV. Does one try to re- build to N^sy? This depends on the willingness of the fishery and management to forgo catch yields during times of increasingly high abundance, possibly for an extended period, so that the population shifts to the higher regime. Regime shifts of long- term stability almost cer- tainly come with a type- IV reference point. In this case, even the closure of the fishery will not gener- ate enough surplus pro- duction to rebuild past the type-III low. Recognizing the existence of such a bar- rier is critical. Presumably, a massive recruitment en- hancement program could be implemented to artifi- cially affect a regime shift. Patience may be the better alternative, using the Nmsy value of the present regime as the management goal o o =1 "O o Q. w Q. 3 c n Population size Figure 8 The relationship of the primary trends in population abun- dance and surplus production associated with the bimodal surplus production trajectory depicted in Figure 7 in which the minimum in surplus production is negative. When surplus production is positive, the population abundance increases. The opposite trend occurs when surplus production is negative. The type-I reference point, the carrying capacity, is a conver- gence. Trends in surplus production and population abundance converge at this point. The type-IV reference point, the point of no return, is a divergence. Trends in surplus production and population abundance diverge at this point. Figure 9 Time series of oyster abundance, by bay region, with the abundance levels associated with types I-IV reference points identified. Regime shifts occurred in 1970 and 1985 (Powell et al., 2009). The 1959 peak is a survey artifact. Total oyster abundance is the cumulative value. Bed groups are defined in Table 1. Bed locations are shown in Figure 1. Reference point legend and symbols are given in order as displayed on the graph, from top to bottom. 144 Fishery Bulletin 107(2) while awaiting the rare sequence of events generating a natural transition to the alternate stable state. Harvest goals Included in Figures 4-7 is an estimated allowable catch as a fraction of the stock. The values of surplus produc- tion given in Figures 4-7 are expressed in numbers, perforce as they are the data source from which the underlying biological relationships are derived. The estimate is provided with some trepidation because the present model does not take into account the differential in growth across the salinity gradient and therefore tends to overestimate the number of animals of market size in the population as a whole. Moreover, the model assumes absolute constancy in the relationship of brood- stock to recruitment. Thus, the model may overestimate the fraction of the stock available for harvest in any given year. The formulation of Klinck et al. (2001) is a preferred option to obtain fishery allocations. Finally, the model consistently predicts a higher harvestable fraction at low abundance than at high abundance. An abettor in this trend may be the reliance of setting larvae more and more on the shell resource at low abundance than on the standing crop of living individuals. However, some portion of this outcome is likely due to an inability to accurately extrapolate the primary biological relation- ships below 0.8 xlO9 animals. Such low abundances have not been observed and therefore the extrapolation is likely to be increasingly in error at lower and lower abundances. We do not give complete credence, therefore, to the proportional increase in harvestable fraction at low abundance indicated by the surplus production tra- jectories depicted in Figures 4-7. From Figure 3 we observe that the range of abun- dances assigned to the various reference points varies little among simulations describing a range of assump- tions about natural mortality and recruitment rate. By contrast, the range of surplus production is pro- digious. Thus, an abundance goal distinguishing an overfished from a sustainable stock, e.g., N , is well constrained, whereas an overfishing definition, e.g., fmsyt is very poorly delimited. Clearly any successful approach to management must minimize the chance that the added mortality by fishing overcomes the in- ertia militating against abundance decline. Further, the uncertainty of the level of surplus production at its minimum and maxima (Fig. 5) necessitates precau- tion as the increased mortality from fishing may be sufficient to stabilize a quasi-stable state at low abun- dance. Both require, for oysters, that fishing mortality be maintained at a small percentage of the natural mortality rate, thereby permitting the inertia of the system to guard against an abundance decline and reducing the chance that a rare population expansion might be prematurely terminated. Even at Nmsy, fishing mortality rate is likely not to exceed 5-10% of the stock (Figs. 4-7). The history of the Delaware Bay fishery provides strong corroboration that removals exceed- ing 15% are not sustainable (Powell et al., 2008) and offers strong evidence that removals below 5% of the stock limit the long-term impact of disease epizootics on abundance. Direct application of the Klinck et al. (2001) model in Delaware Bay has routinely returned values in the range of 1-3%. In addition, Powell and Klinck (2007) discuss the impact of fishing on the shell resource and the degradation of the shell beds upon which the population depends for its existence. That analysis independently argues for fishing mortality rates distinctly below the predisease mortality rate, at approximatly 10%. It is noteworthy that allowable fishing mortality rates <10% of the stock are more similar to the mortality rates of the longest-lived bivalves, such as geoducks and ocean quahogs (e.g., Bradbury and Tagart, 2000; NEFSC2), than other species with life spans of the same order as the oyster, emphasizing the fact that oysters in the Mid-Atlantic region are much more akin in their population dynamics to long-lived k-selected species than to short lived r-selected ones.3 Low recruitment significantly restricts the ambit of the oyster’s popula- tion dynamics and significantly constrains allowable fishing mortality rates over a wide range of abundance values. A perusal of the broodstock-recruitment curve (Fig. 7 in Powell et al., 2009) shows that recruitment rate typically falls within the range of 0.25 to 1 spat per adult animal per year. Both this recruitment level and the <10%-per-year natural mortality rate is consistent with theoretical predisease generation times that likely exceeded 10 years (Mann and Powell, 2007) and the fact that reproduction continues to be consistent with an animal characterized by longer generation times. Conclusions The oyster population in Delaware Bay exhibits a popu- lation dynamics that is not normally described in com- mercial species. One reason is the presence of distinct and dynamically stable multiple stable points delimited by temporally rapid regime shifts. The result of this complexity is a series of reference points identified by the trajectory of surplus production, which departs dramati- cally from the simple Schaefer curve (e.g., Zabel et al., 2003). We define four reference point types in terms of surplus production, its derivative, and the rate of change of this derivative (Table 2). In Delaware Bay, the surplus production trajectory likely manifests two stable points and the carrying capacities associated with them and these agree relatively well with the observed stable 2 NEFSC (Northeast Fisheries Science Center). 2001. 33rd northeast regional stock assessment workshop (33rd SAW): Stock assessment review committee (SARC) consensus summary of assessments. NMFS NEFSC Ref. Doc. 01-18, 281 p. 3 Gulf of Mexico conditions with rapid growth (Ingle and Dawson, 1952; Butler, 1953; Hayes and Menzel, 1981) and multiple spawns per year (Hopkins, 1954; Hayes and Menzel, 1981; Choi et al., 1993, 1994) are examples of C. virginica under more r-selected conditions. Powell et al.: Multiple stable reference points in oyster populations 145 states in the population time series (Fig. 9). For each of these type-I reference points, a maximum in surplus production also exists. The presence of two stable states assures a type-III reference point that is a measure of the ease of transition between the two stable states and provides information on the likelihood that management can artificially impose a transition. In Delaware Bay, the type-III surplus production value may be negative. In this case, a type-IV reference point exists, a point- of-no-return. If the type-III reference point is positive, a quasi-stable state exists at low abundance that can be stabilized by overfishing. The existence of a positive type-III reference point imposes a particular conundrum to management in that rebuilding requires a reduction in fishery yield as abundance increases over a substan- tive abundance range. The simulations show the uncertainty imposed by the limitations on accurate knowledge of the biological relationships. One noteworthy observation is that the location of the reference points undefined by a specific surplus production value (e.g., St=0), namely types II and III, are relatively stable in position with respect to population abundance over a wide range of uncertain- ties in recruitment and mortality rates (Table 2). The surplus production values associated with these refer- ence points are much more uncertain (Table 2). Thus, location is much better known than scale. As recom- mended by Beverton et al. (1984), different models are likely to be needed for short-term catch forecasts and estimation of abundance goals. We describe reference points in the context of multiple stable states. The simplicity of the Bmsy-K couple so emphasized in fisheries management fails when mul- tiple stable states exist. That they may often exist is now well considered, although not yet inculcated into the oracle of fisheries management. Multiple stable points assure 1) that a type-III reference point exists, 2) that this point will impede the attainment of impru- dently formulated rebuilding goals, 3) that a type-IV point-of-no-return may exist that establishes a barrier to rebuilding, as well as imposing the conditions at high abundance necessary for stock collapse, and 4) that a carrying capacity may exist at abundances well below historically high abundances and well below the simplistic promulgation of Bmsy as half the carrying capacity established by the higher stable state. Use of the latter may impose impossible requirements for re- building a stock because the promulgated goal exceeds the carrying capacity for the controlling regime. Acknowledgments We recognize the many people who contributed over the years to the collection of the 54 years of survey data analyzed in this report, with particular recognition of the contributions by H. Haskin, D. Kunkle, and B. Richards. We appreciate the many suggestions on con- tent provided by S. Ford and D. Bushek. The study was funded by an appropriation from the State of New Jersey to the Haskin Shellfish Research Laboratory, Rutgers University, and authorized by the Oyster Industry Sci- ence Steering Committee, a standing committee of the Delaware Bay Section of the Shell Fisheries Council of New Jersey. Literature cited Abbe, G. R. 1988. Population structure of the American oyster, Crassostrea virginica, on an oyster bar in central Chesa- peake Bay: Changes associated with shell planting and increased recruitment. J. Shellfish Res. 7:33-40. Anonymous. 1996. Magnuson-Stevens Fishery Conservation and Man- agement Act. NOAA Tech. Memo. NMFS-F/SPO-23, 121 p. Beverton, R. J. H., J. G. Cooke, J. B. Csirke, R. W. Doyle, G. Hempel, S. J. Holt, A. D. MacCall, D. J. Policansky, J. Roughgarden, J. G. Shepherd, M. P. Sissenwine, and P. H. Wiebe. 1984. Dynamics of single species group report. In Exploi- tation of marine communities (R. M. May, ed.), p. 13— 58. Dahlem Konferenzen, Springer-Verlag, Berlin. Bradbury, A., and J. V. Tagart. 2000. Modeling geoduck, Panopea abrupta (Conrad, 1849) population dynamics II. Natural mortality and equilib- rium yield. J. Shellfish Res. 19:63-70. Breitburg, D. L., L. D. Coen, M. W. Luckenbach, R. Mann, M. Posey, and J. A. Wesson. 2000. Oyster reef restoration: Convergence of harvest and conservation strategies. J. Shellfish Res. 19:371- 377. Burreson, E. M., and L. M. Ragone Calvo 1996. Epizootiology of Perkinsus marinus disease of oys- ters in Chesapeake Bay, with emphasis on data since 1985. J. Shellfish Res. 15:17-34. Butler, P. A. 1953. Oyster growth as affected by latitudinal tempera- ture gradients. Commer. Fish Rev. 15:7-12. Caffey, H. M. 1985. Spatial and temporal variation in settlement and recruitment of intertidal barnacles. Ecol. Monogr. 55:313-332. Choi, K-S., D. H. Lewis, E. N. Powell, and S. M. Ray. 1993. Quantitative measurement of reproductive output in the American oyster, Crassostrea virginica (Gmelin), using an enzyme-linked immunosorbent assay (ELISA). Aquacult. Fish. Manag. 24:299-322. Choi, K-S., E. N. Powell, D. H. Lewis, and S. M. Ray. 1994. Instantaneous reproductive effort in female Ameri- can oysters, Crassostrea virginica, measured by a new immunoprecipitation assay. Biol. Bull. (Woods Hole) 186:41-61. Clark, W. G. 1999. Effects of an erroneous natural mortality rate on a simple age-structured stock assessment. Can. J. Fish. Aquat. Sci. 56:1721-1731. Collie, J. S., K. Richardson, and J. H. Steele. 2004. Regime shifts: Can ecological theory illuminate the mechanisms? Prog. Oceanogr. 60:281-302. FAO (Food and Agriculture Organization of the United Nations). 1995. Technical consultation on the precautionary approach to capture fisheries. Precautionary approach 146 Fishery Bulletin 107(2) to fisheries. Part 1: Guidelines on the precaution- ary approach to capture fisheries and species intro- ductions. FAO Fisheries Tech. Pap. no. 350, part A, 52 p. FAO, Rome. Francis, R. I. C. C., and R. Shotton. 1997. “Risk” in fisheries management: A review. Can. J. Fish. Aquat. Sci. 54:1699-1715. Frank, S. A. 1996. Models of parasite virulence. Q. Rev. Biol. 71:37-78. Frechette, M., and E. Bourget. 1985. Food-limited growth of Mytilus edulis L. in relation to the benthic boundary layer. Can. J. Fish. Aquat. Sci. 42:1166-1170. Frechette, M., and D. Lefaivre. 1990. Discriminating between food and space limita- tion in benthic suspension feeders using self-thinning relationships. Mar. Ecol. Prog. Ser. 65:15-23. Godfray, H. C. J., and C. J. Briggs. 1995. The population dynamics of pathogens that control insect outbreaks. J. Theor. Biol. 176:125-136. Gray, J. S. 1977. The stability of benthic ecosystems. Helgol. Wiss. Meeresunters. 30:427-444. Haddon, M. 2001. Modelling and quantitative methods in fisheries, 406 p. Chapman and Hall, Boca Raton, FL. Hancock, D. A. 1973. The relationship between stock and recruitment in exploited invertebrates. Rapp. P-v. Reun Cons. Int. Explor. Mer 164:113-131. Hasiboder, G., C. Dye, and J. Carpenter. 1992. Mathematical modelling and theory for esti- mating the base reproduction number of canine leishmaniasis. Parasitology 105:43-53. Hassell, M. P., R. C. Anderson, J. E. Cohen, B. Cujetanovic, A. P. Dobson, D. E. Gill, J. C. Holmes, R. M. May, T. McKeown, and M. S. Pereira. 1982. Impact of infectious diseases on host populations. In Population biology of infectious diseases (R. M. Anderson, and R. M. May, eds.), p. 15-35. Dahlem Konferenzen, Springer-Verlag, New York, NY. Haven, D. S., and J. P. Whitcomb. 1983. The origin and extent of oyster reefs in the James River, Virginia. J. Shellfish Res. 3:141-151. Hayes, P. F., and R. W. Menzel. 1981. The reproductive cycle of early setting Crassostrea virginica (Gmelin) in the northern Gulf of Mexico, and its implications for population recruitment. Biol. Bull. (Woods Hole) 160:80-88. Heesterbeek, J. A. P., and H. G. Roberts. 1995. Mathematical models for microparasites of wildlife. In Ecology of infectious diseases in natural populations (B. T. Grenfell, and A. P. Dobson, eds.), p. 90-122. Cambridge Univ. Press, Cambridge, UK. Hethcote, H. W., and P. van den Driessche. 1995. An SIS epidemic model with variable population size and a delay. J. Math. Biol. 34:177-194. Hilborn, R., and C. J. Walters. 1992. Quantitative fisheries stock assessment. Choice, dynamics and uncertainty, 570 p. Chapman and Hall, New York, NY. Holmes, J. C. 1982. Impact of infectious disease agents on the population growth and geographical distributions of animals. In Population biology of infectious diseases (R. M. Anderson, and R. M. May, eds.), p. 37-51. Dahlem Konferenzen, Springer-Verlag, New York, NY. Hopkins, S. H. 1954. Oyster setting on the Gulf coast. Proc. Natl. Shellfish. Assoc. 45:52—55. Imeson, R. J., J. C. J. M. van den Bergh, and J. Hoekstra. 2002. Integrated models of fisheries management and policy. Environ. Model. Assess. 7:259-271. Ingle, R. M., and C. E. Dawson Jr. 1950. Variation in salinity and its relation to the Florida oyster. Part One: Salinity variations in Apalachicola Bay. Proc. Natl. Shellfish. Assoc, p. 6-19. Johnson, L. 1994. Pattern and process in ecological systems: A step in the development of a general ecological theory. Can. J. Fish. Aquat. Sci. 51:226-246. Jordan, S. J., K. N. Greenhawk, C. B. McCollough, J. Vanisko, and M. L. Homer 2002. Oyster biomass, abundance, and harvest in north- ern Chesapeake Bay: Trends and forecasts. J. Shellfish Res. 21:733-741. Keough, M. J., and B. J. Downes. 1982. Recruitment of marine invertebrates: The role of active larval choices and early mortality. Oecologia (Berl.) 54:348-352. Kermack, W. O., and A. G. McKendrick. 1991. Contributions to the mathematical theory of epi- demics— I. Bull. Math. Biol. 53:33—55. Klinck, J. M., E. N. Powell, J. N. Kraeuter, S. E. Ford, and K. A. Ashton-Alcox. 2001. A fisheries model for managing the oyster fish- ery during times of disease. J. Shellfish Res. 20:977- 989. Knowlton, N. 2004. Multiple “stable” states and the conservation of marine ecosystems. Prog. Oceanogr. 60:387-396. Leffler, M. 2002. Crisis and controversy. Does the bay need a new oyster? Chesapeake Quart. 1:2-9. Lipton, D. W., and I. E. Strand. 1992. Effect of stock size and regulations on fishing indus- try cost and structure: The surf clam industry. Am. J. Agr. Econ. 74:197-208. Mackinson, S., U. R. Sumaila, and T. J. Pitcher. 1997. Bioeconomics and catchability: Fish and fishers behaviour during stock collapse. Fish. Res. 31:11— 17. Mangel, M., and C. Tier. 1994. Four facts every conservation biologist should know about persistence. Ecology 75:607-614. Mann, R. 2000. Restoring the oyster reef communities in the Chesapeake Bay: A commentary. J. Shellfish Res. 19:335-339. Mann, R., E. M. Burreson, and P. K. Baker. 1991. The decline of the Virginia oyster fishery in Chesapeake Bay: Considerations for the introduction of a non-endemic species, Crassostrea gigas (Thunberg, 1793). J. Shellfish Res. 10:379-388. Mann, R., and E. N. Powell. 2007. Why oyster restoration goals in the Chesapeake Bay are not and probably cannot be achieved. J. Shellfish Res. 26:905-917. May, R. M., J. R. Beddington, J. W. Horwood, and J. G. Shepherd. 1978. Exploiting natural populations in an uncertain world. Math. Biosci. 42:219-252. Powell et al. : Multiple stable reference points in oyster populations 147 McCann, K. S., L. W. Botsford, and A. Hasting. 2003. Differential response of marine populations to cli- mate forcing. Can. J. Fish. Aquat. Sci. 60:971—985. Peterson, C. H. 1984. Does a rigorous criterion for environmental identity preclude the existence of multiple stable points? Am. Nat. 124:127-133. Powell, E. N., and K. A. Ashton-Alcox. 2004. A comparison between a suction dredge and a tra- ditional oyster dredge in the transplantation of oysters in Delaware Bay. J. Shellfish Res. 23:803-823. Powell, E. N., K. A. Ashton-Alcox, S. E. Banta, and A. J. Bonner. 2001. Impact of repeated dredging on a Delaware Bay oyster reef. J. Shellfish Res. 20:961-975. Powell, E. N., K. A. Ashton-Alcox, J. A. Dobarro, M. Cummings, and S. E. Banta. 2002. The inherent efficiency of oyster dredges in survey mode. J. Shellfish Res. 21:691-695. Powell, E. N., K. A. Ashton-Alcox, and J. N. Kraeuter. 2007. Reevaluation of eastern oyster dredge efficiency in survey mode: Application in stock assessment. N. Am. J. Fish. Manag. 27:492-511. Powell, E. N., K. A. Ashton-Alcox, J. N. Kraeuter, S. E. Ford, and D. Bushek. 2008. Long-term trends in oyster population dynam- ics in Delaware Bay: Regime shifts and response to disease. J. Shellfish Res. 27:729-755. Powell, E. N., H. Cummins, R. J. Stanton Jr., and G. Staff. 1984. Estimation of the size of molluscan larval settle- ment using the death assemblage. Estuarine Coastal Shelf Sci. 18:367-384. Powell, E. N., and J. M. Klinck. 2007. Is oyster shell a sustainable estuarine resource? J. Shellfish Res. 26:181-194. Powell, E. N., J. M. Klinck, K. A. Ashton-Alcox, J. N. Kraeuter. 2009. Multiple stable reference points in oyster popu- lations: biological relationships for the eastern oyster (Crassostrea virginica ) in Delaware Bay. Fish. Bull. 107:109-132. Powell, E. N., J. M. Klinck, E. E. Hofmann, E. A. Wilson-Ormond, andM. S. Ellis. 1995. Modeling oyster populations. V. Declining phyto- plankton stocks and the population dynamics of Ameri- can oyster (Crassostrea virginica ) populations. Fish. Res. 24:199-222. Powell, E. N., J. N. Kraeuter, and K. A. Ashton-Alcox. 2006. How long does oyster shell last on an oyster reef? Estuarine Coastal Shelf Sci. 69:531-542. Ragone Calvo, L. M., R. L. Wetzel, and E. M. Burreson 2001. Development and verification of a model for the population dynamics of the protistan parasite, Per- kinsus marinus, within its host, the eastern oyster, Crassostrea virginica, in Chesapeake Bay. J. Shellfish Res. 20:231-241. Redner, S. 2001. A guide to first-passage processes, 312 p. Cam- bridge Univ. Press, Cambridge, U.K. Restrepo, V. R., G. G. Thompson, P. M. Mace, W. K. Gabriel, L. L. Low, A. D. MacCall, R. D. Methot, J. E. Powers, B. L. Taylor, P. R. Wade, and J. F. Witzig. 1998. Technical guidance on the use of precautionary approaches to implementing National Standard 1 of the Magnuson/Stevens Fishery Conservation and Man- agement Act. NOAA Tech. Memo. NMFS-F/SPO-31, 54 p. Rice, J. 2001. Implications of variability on many time scales for scientific advice on sustainable management of living marine resources. Prog. Oceanogr. 49:189-209. Ricker, W. E. 1975. Computation and interpretation of biological sta- tistics of fish populations. Bull. Fish. Res. Board Can. 191:1-382. Rothschild, B. J., J. S. Ault, P. Goulletquer, and M. Heral 1994. Decline of the Chesapeake Bay oyster populations: A century of habitat destruction and overfishing. Mar. Ecol. Prog. Ser. 111:29-39. Rothschild, B. J., C. Chen, and R. G. Lough. 2005. Managing fish stocks under climate uncertainty. ICES J. Mar. Sci. 62:1531-1541. Rothschild, B. J., and A. J. Mullen. 1985. The information content of stock-and-recruitment data and non-parametric classification. J. Cons. Int. Explor. Mer. 42:116-124. Soniat, T. M., E. N. Powell, E. E. Hofmann, and J. M. Klinck. 1998. Understanding the success and failure of oyster populations: The importance of sampled variables and sample timing. J. Shellfish Res. 17:1149-1165. Soniat, T. M., E. N. Powell, J. M. Klinck, and E. E. Hofmann. In press. Understanding the success and failure of oyster populations: Climatic cycles and Perkinsus marinus. Int. J. Earth Sci. Swinton, J., and R. M. Anderson. 1995. Model frameworks for plant-pathogen inter- actions. In Ecology of infectious diseases in natural populations (B. T. Grenfell, and A. P. Dobson, eds.), p. 280-294. Cambridge Univ. Press, Cambridge, U.K. Zabel, R. W„ C. J. Harvey, S. L. Katz, T. P. Good, and P. S. Levin. 2003. Ecologically sustainable yield. Am. Sci. 91:150- 157. 148 Abstract — Estimation of individual egg production (realized fecundity) is a key step either to understand the stock and recruit relationship or to carry out fisheries-independ- ent assessment of spawning stock biomass using egg production meth- ods. Many fish are highly fecund and their ovaries may weigh over a kilogram; therefore the work time can be consuming and require large quantities of toxic fixative. Recently it has been shown for Atlantic cod (Gadus morhua) that image analy- sis can automate fecundity determi- nation using a power equation that links follicles per gram ovary to the mean vitellogenic follicular diam- eter (the autodiametric method). In this article we demonstrate the precision of the autodiametric method applied to a range of species with dif- ferent spawning strategies during maturation and spawning. A new method using a solid displacement pipette to remove quantitative fecun- dity samples (25, 50, 100, and 200 mil- ligram [mg]) is evaluated, as are the underlying assumptions to effectively fix and subsample the ovary. Finally, we demonstrate the interpretation of dispersed formaldehyde-fixed ovarian samples (whole mounts) to assess the presence of atretic and postovulatory follicles to replace labor intensive his- tology. These results can be used to estimate down regulation (production of atretic follicles) of fecundity during maturation. Manuscript submitted 2 May 2008. Manuscript accepted 3 October 2008. Fish. Bull. 107:148-164 (2009). The views and opinions expressed or implied in this article are those of the author and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Advances in methods for determining fecundity: application of the new methods to some marine fishes Peter R. Witthames (contact author)1 Lorraine N. Greenwood2 Anders Thorsen2 Rosario Dominguez'1 Hilario Murua Maria Korta3 Francisco Saborido-Rey4 Olav S. Kjesbu2 Email address for contact author: fecund-fish@tiscali.co.uk. 1 Centre for Environment, Fisheries & Aquaculture Science (Cefas) Lowestoft, Suffolk NR33 OHT, United Kingdom Present address: Fecund Fish Consultancy 40 Plumtrees Lowestoft, Suffolk NR32 3JH, United Kingdom 2 Institute of Marine Research (IMR) P.O. Box 1870 Nordnes, N-5817 Bergen, Norway 3 AZTI Tecnalia Herrera Kaia Portualde z/g 20110 Pasaia, Spain 4 Institute of Marine Research (CSIC) Eduardo Cabello 6, 36208, Vigo, Spain Research on population fecundity (total egg production) has two impor- tant applications in the management of renewable marine or freshwater fish resources. Perhaps the most impor- tant is to understand the relationship between spawning stock biomass and recruitment because it is increasingly clear that the assumption of direct pro- portionality (Beverton and Holt, 1957) is not correct (Marshall et al., 1998; Witthames and Marshall, 2008). The link between spawning stock biomass and recruitment varies according to total egg production that in turn is dependent, not only on the length fre- quency of spawning adults, but also on body weight at length (Marshall et al., 1998, 1999). In summary the stock and recruitment relationship became stronger when stock was expressed as a product of population length fre- quency and fecundity at length (Mar- shall et al., 1998). A second application for fecundity information is to estimate spawn- ing stock biomass independently of data collected from commercial fish- eries (Parker, 1980; Lockwood et al., 1981; Lo et al., 1992). In addition to the fecundity count, information is required on a range of parameters associated with the development of fecundity, such as follicular diameter and frequency distribution, spawning rates, and individual realized fecun- dity, either in one or multiple batches shed during one or more spawning events. In this article we prefer the follow- ing definitions for fecundity (Hunter et al., 1992). Thus the developing fecundity (standing stock of fecun- dity referred to as “fecundity” [F] ) includes follicles containing cortical alveoli (Khoo, 1979) and, in later development, yolk granules (Hunter et al., 1992) but excludes precursor cells such as previtellogenic follicles (PVFs) or oogonia. Relative fecundity (Fbw) is the fecundity divided by the total fish weight. We use the term “follicle” to refer to the oocyte and its nurturing follicular layers (Tyler and Sumpter, 1996) during all phases of development from precursor cells to residual postovulatory follicles (POFs) that indicate previous spawning or egg release events. The species in this study are of in- terest because they represent three extremes in spawning strategy (Mu- Witthames et al.: Advances in methods for determining fecundity in marine fishes 149 rua and Saborido-Rey, 2003): 1) group synchronous determinate total spawners (Atlantic herring [Clu- pea harengus], deep water redfish [Sebastes mentella ] also known as "beaked redfish [FAO, Fisheries and Aquaculture Dept., www.fao.org/fishery/statistics/ programme/3.1.1. Accessed Jan., 2009], and golden redfish [Sebastes marinus ]); 2) group synchronous de- terminate batch spawners (Atlantic cod [Gadus morhua ] and European plaice [Pleuronectes platessa ]); and 3) asynchronous types (European hake [Merluccius mer- luccis\ and Atlantic mackerel [Scomber scombrus ]) that may not be determinate (Greer-Walker et al., 1994). Fecundity in the first two groups included all follicles in the advanced mode to the right of a gap in the fol- licular size frequency (Hunter et al., 1992) whereas in the latter case the follicular distribution is continuous. Although fecundity may be enhanced during maturation in asynchronous spawning types (indeterminate spawn- ing strategy), it is of practical and theoretical value to study fecundity proliferation whatever classification is applied to the spawning process. Recent work has shown that not all the fecundity develops into eggs (realized fecundity) and follicular atresia may account for a substantial part of the fe- cundity in a process referred to as down regulation (see reviews Murua et al., 2003; Thorsen et al., 2006; Kjesbu and Witthames, 2007). In addition, it is also important to differentiate whether an individual female has en- tered the spawning cycle, thus reducing her fecundity, and how long a POF persists to indicate a previous spawning event (Hunter and Macewicz, 1985a). The lat- ter information is used to assess whether a female still contains her full complement of oocytes for application of the annual egg production method to assess spawn- ing stock biomass applied to fish with a determinate spawning strategy (Armstrong et al., 2001). To date no single approach has been successful in quantifying follicular stages associated with fecundity development and regression and each has one or more disadvantages. Fish that are very fecund, perhaps con- taining ovaries weighing more than a kilogram and with millions of follicles, will have to be subsampled for fecundity estimation. In this case, quantitative histolog- ical methods (Emerson et al., 1990) requiring sections of the whole ovary are not feasible — meaning only relative proportions of each follicular class can be measured (An- dersen, 2003). This approach, however, needs additional information on the fecundity count preferably coupled with measurement of follicular size frequency to exclude smaller PVFs that are not committed to maturation in the current reproductive year. Although it is feasible to release follicles by digesting the ovary in strong acid solutions (either Gilson’s fluid (Simpson, 1951) or a less toxic nitric acid formulation (Friedland et al., 2005)), such media have several adverse consequences. These consequences include 1) considerable follicular shrink- age (Witthames and Greer-Walker, 1987), 2) likely loss of atretic follicles and POFs (Klibansky and Juanes, 2007), and 3) incompatibility with histological methods (Hunter and Macewicz, 2003). In view of the need to identify fecundity based on follicular size, there is a need to measure large numbers of follicles greater than a specified lower size limit even if the ovary is subsam- pled using the gravimetric method (Bagenal and Braum, 1968). Manual measurement of follicular size frequency, even using video technology, is just too demanding on manual labor unless there is some way of automating the collection of data. Although an automatic particle analyzer can provide such data (Witthames and Greer- Walker, 1987), the method requires large quantities of Gilson’s fluid and is subject to all the problems listed above. More recently image analysis methods have been adopted to automate collection of size frequency data in Atlantic cod ( Gadus morhua) (Thorsen and Kjesbu, 2001; Klibansky and Juanes, 2008). In each case the mean fecundity (the independent variable) can be used to estimate the number of follicles per gram (g) of ovary by fitting a power relationship based on a calibration from a data set containing the two variables (the auto- diametric method). Fecundity is then determined by raising the number of follicles per g of ovary by the ovarian weight. Although an alternative image analysis method applied to American shad ( Alosa sapidissima) (Friedland et al., 2005) has advantages as a cost ef- fective method to estimate fecundity, it also has two significant drawbacks: 1) relatively low resolution, and, 2) it uses acid hydrolysis to separate follicles. Thus, the autodiametric method has more general utility because it uses neutral buffered formaldehyde solution (NBF) to fix tissue that is fully compatible with histology. Also Hunter et al. (1992) studying Dover sole ( Microstomas pacificus) and Oskarsson et al. (2002) studying Atlantic herring ( Clupea harengus ) have shown it possible to identify atretic follicles in NBF-fixed dispersed ovarian samples (whole mounts) suggesting it might be pos- sible to also estimate numbers of different follicular classes. Accordingly, our first objective is to report on the utility and precision of the autodiametric method to determine fecundity in several species including At- lantic cod, European hake, Atlantic herring, Atlantic mackerel, redfish species (deep water redfish and golden redfish), and European plaice. To emphasize the utility of the method several laboratories 1) AZTI [A] (Pasaia, Spain), 2) Cefas [B]( Lowestoft, UK), 3) CSIC [C] (Vigo, Spain), and 4) IMR [D] (Bergen, Norway) used different configurations of image analysis equipment. In order to complete this work, four other objectives were identi- fied linked to the application of fecundity determina- tion using Atlantic cod as the main example and to a lesser extent European hake: 1) ovarian sampling, and follicular homogeneity, 2) evaluate three stains (eosin, rose bengal, and periodic acid-schiff [PAS]) to improve the accuracy of follicular size measurement and count- ing in relation to the autodiametric method, 3) compare interpretation of NBF-fixed whole mounts with respect to histology to assess maturity, spawning status and quantify the standing stock of atretic follicles, and 4) consideration was also given to the effect of ovarian maturation on down regulation of fecundity in Atlantic 150 Fishery Bulletin 107(2) Table 1 Details of the collection date, source, and maturity stage of wild fish (Atlantic cod [Gadus morhua ], haddock [Melanogrammus aeglefinus], European hake [Merluccius merluccius], Atlantic herring [Clupea harengus], Atlantic mackerel [Scomber scotnbrus], European plaice [Pleuronectes platessa ], common sole [Solea solea], redfish (deep water redfish [Sebastes mentella ] or golden redfish [Sebastes marinus]) used in the study. The samples where taken from commercial or research vessel catches. Collections made in 1995 were used for the study of fecundity down regulation and the later collections from 1998 onward for the study of fecundity methods. Atlantic Cod Haddock European hake Atlantic herring Atlantic mackerel European plaice Common sole Redfish Date Jan-Mar 1995, 2003-04, 2007 Feb 2007 Mar 2003 Jan 1998 Mar 2004 Jan-Mar 1995, 2000, 2007 Jan-Feb 1995 Sep-Nov 2000-2001 Source Lofoten Andenes Norway North and Irish Sea Irish Sea Galicia Biscay Celtic Sea Norwegian- spring spawning stock Western Atlantic stock Irish Sea Irish Sea Iceland Irminger Sea Maturity stage Prespawning and spawning Pre- spawning Pre- spawning Pre- spawning Pre- spawning Pre- spawning Pre- spawning Pre- spawning and to near spawning spent cod, European plaice, and common sole ( Solea solea) and the implications for estimating total fecundity prior to the start of spawning. Materials and methods Ovarian sampling, follicle measurement equipment, and homogeneity Ovarian samples were collected by the four institutes working on two or more of the following species Atlantic cod, Atlantic herring, European hake, Atlantic mackerel, European plaice, and redfish (including deep water redfish and golden redfish) for studies unrelated to this paper (Table 1). Biological information was taken from each fish, but only the information related to this method development (ovarian mass and maturity stage) is used here. Only active ovaries (Hunter et al., 1992) were selected and weighed to better than 2% of their mass, either with a motion-compensated balance (POLS Elec- tronics, Isafjordur, Iceland) when sampled at sea or with a standard balance when on shore. Fish that contained many ovulated eggs (caught in the act of spawning) were not used in the autodiametric calibration because they show a heterogeneous distribution (Witthames, 2003). In each case ovaries or ovarian subsamples were fixed in a minimum of two volumes of NBF for a minimum of 14 days. Quantitative subsamples were taken by one of two methods: 1) from the fresh unfixed ovary (Atlan- tic cod, Atlantic mackerel) immediately after capture at sea using a Wiretrol II pipette (Drummond Scien- tific, Broomall, PA), or 2) from the fixed ovary in the laboratory with a scalpel (Atlantic cod, European hake, European plaice, and redfish). The Wiretrol II pipette consists of a Teflon-tipped stainless steel piston within a graduated glass tube with a 1 or 2 millimetre (mm) bore that can remove 26 and 54, or 106 and 212 milligrams (mg), respectively, of tissue when inserted through the ovarian tunica. Image analysis hardware and software, including the camera resolution and light intensities used by each institute to measure follicular diameter and circularity, varied (Table 2). The follicle data were analyzed with Microsoft Excel to calculate follicular mean, standard deviation, and leading cohort (Lc) (defined as the mean of largest 10% of follicles measured). PVFs were exclud- ed from the fecundity count and frequency distribution based on a minimum follicular diameter of 150 and 250 pm in European hake, and Atlantic cod, respectively (Kjesbu, 1991; Murua and Motos, 2006). In the case of Atlantic mackerel there were no published data avail- able, so a diameter of 185 pm was used based on our observation of the smallest follicles containing cortical alveoli and a publication focusing on Atlantic mackerel fecundity determinacy (Greer-Walker et al., 1994). In all other species, where the follicular frequency was not continuous, only dark yolk-bearing follicles in the leading mode were included in the fecundity count. Fol- licles were disaggregated from the ovarian sample by sucking them in and out of a Pasteur pipette (Thorsen and Kjesbu, 2001) prior to spreading them out in a counting chamber as a single layer completely covered by water. If small clumps of follicles remained they were measured manually as discrete follicles, whereas larger clumps were separated and exposed to more pressure washing by the pipette or a garden spray (Institutes A and C). All the follicles were counted in the subsam- ples but the method for collecting the follicular meas- urements differed according to each institute’s image Witthames et al. : Advances in methods for determining fecundity in marine fishes 151 Table 2 Details of the image analysis software PPilkington Image Analysis Systems,2 Shareware available at rsb.info.nih.gov/ij/down- load.html), operating systems, camera, light source, microscope, resolution, grey scale, and staining methods used by each insti- tute: AZTI, Cefas, CSIC, and IMR to analyze whole mounts prepared from each species: Atlantic cod ( Gadus morhua ), European hake (Merluccius merluccius , Atlantic herring ( Clupea harengus ), Atlantic mackerel ( Scomber scombrus), European plaice ( Pleu - ronectes platessa), redfish (deep water redfish (Sebastes mentella ) or golden redfish ( Sebastes marinus). Detail AZTI [A] Cefas [B] CSIC [C] IMR [D] Software Visilog 6.11 Myrmica1 automatic or semi automatic operation QWin (Leica Imaging System) Image SXM v. 1.772 Operating system PC Windows PC Windows PC Windows Mac OS X Camera Camedia-4040 Zoom Pulnix TMC-1000-CL Leica IC A JVC TK-1070E Light source 3100 High light Olympus 3100 Olympus High light Leica MZ6 SZX-ILLB200 Microscope Olympus SZX12 Olympus SZX12 Leica MZ6 Olympus SZX12 Resolution (pm/pixel) 5 (hake) 19.3 (plaice), 10.0 (cod), 4.6 (mackerel) 21.9-27.6 (redfish) 7.02-22.43 (hake) 14.7 (cod) 14.7 (herring) Gray scale (255 saturated) 230 175 90 205 Staining method Rose bengal (hake) PAS (mackerel and cod) Eosin (plaice) Rose bengal (hake) Unstained redfish Unstained analysis system and the size of the follicles comprising the fecundity. Institutes A and C split the sample and took a separate image of each aliquot to ensure that the follicles were evenly spread without overlap in a container that was completely covered within the field of view. Institute B spread the sample in a counting chamber 70 mm long and either 4, 7, or 10 mm wide for Atlantic mackerel, Atlantic cod, and European plaice, respectively. The three widths of counting chamber were used so that three magnifications (Table 2) could be used while still displaying the full width of the cham- ber. Above the sample the counting chamber tapered outwards in v-shaped profile leading to an upper liq- uid surface of 25 mm creating a flat meniscus over the channel holding the follicles. Myrmica software (Pilkington image analysis systems, Lindfield, West Sussex, UK) was used to create and archive a series of images and overlays along the horizontal axis of the sample container showing all measurements overlaying the follicles measured. Individual follicles stained PAS were measured in this counting chamber both manually and by image analysis to establish the accuracy of size measurement by image analysis. Institute D working with Atlantic cod and Atlantic herring used a method described previously (Thorsen and Kjesbu, 2001). In order to investigate whether fixing the ovary in sample aliquots or whole in the tunica affected mean follicular diameter (Df , pm), circularity, and fecundity per gram of ovary (Fow), replicate samples were taken by pipette and scalpel, respectively, from the central part of the same ovary from Atlantic cod, Atlantic haddock, and European plaice (Table 1). These samples were collected from fish caught in the Irish Sea during February 2007 (Table 1) and fixed for between 63 and 91 days in 1.7 to 9 times their volume of NBF before image analysis (Table 2). Circularity, a function of follicular shape, was measured according to the following equation: Circularity = 4ji (area / perimeter)2 . (1) Homogeneity of Df and Fow were studied in Atlantic cod and European hake ovaries to investigate whether a sample from the center of the ovarian mass was sig- nificantly different from samples taken at the extreme ends of each ovary or between the pair of ovaries. Stain evaluation Ovarian tissue (whole mounts) stained by the three routines (Table 2) was compared with nonstained tissue in order to improve the identification and measure- ment of developing (cortical alveoli and vitellogenic) and regressing (postovulatory and atretic) follicles. Non- stained tissue was prepared for analysis as described by Thorsen and Kjesbu (2001), and two of the staining methods applied water soluble 1% eosin or 0.02% rose bengal weight to volume (w/v) dissolved in NBF to color the follicles. A third staining method used the PAS reaction, previously applied to stain cortical alveoli fol- licles (Greer-Walker et al., 1994). In this procedure the concentration of PAS reagent was 0.1% and 15% w/v, respectively, compared to the histological procedure in order to minimize shrinkage of follicles. Nonbound stain was removed from the tissue subsamples after 152 Fishery Bulletin 107(2) staining by washing through mesh sieves that retained all follicles larger than 125 pm using either 1:3 glyc- erine:water (McBride and Thurman, 2003), 0.9% w/v sodium chloride (Ramsay and Witthames, 1996), or clean water. Replacement of the fixative used for stor- age by saline or water did not affect the size of fol- licles during subsequent storage for 5 days at 0-5°C. Comparison of whole-mount method with histological method In order to study the morphology of POFs immediately after ovulation, and during advanced regression, ovarian samples were taken from trawl caught Atlantic cod taken from the Irish Sea (Table 1) during the spawning season. In some cases the fish (n = 10) were producing copious quantities of ovulated eggs and were expected to con- tain newly produced POFs created simultaneously with ovulation. After fixation in NBF the whole mounts were examined both unstained and after PAS staining to color both the oocyte chorion and the basement membrane between the granulosa and thecal layers of the follicle. The size frequency of the residual vitellogenic follicles and POFs were also measured at this time. Normal vitel- logenic and POFs were tentatively identified in the above preparations based mainly on their shape but also on their internal structure revealed as irregular blotches or shading (Hunter and Macewicz, 1985a, 1985b; Hunter et al., 1992). One fish was chosen from this group because it contained not only large POFs but also illustrated pre- vious spawning activity based on large numbers of small POFs assumed to come from previous ovulation events. Examples of tentatively identified follicular classes were removed from the whole mount and processed into PAS Mallory trichrome stained 2-hydroxyethyl methacrylate (Technovit® 7100 Kulzer GmbH, Wehrheim, Germany) sections (Witthames and Greer- Walker, 1995) in order to compare the accuracy of the identification. Alpha atretic follicles (Hunter and Macewicz, 1985b) were identified in biopsy samples taken with a Pipelle de Cornier® (Prodimed, Neuilly En Thelle, Picardie, France), a flexible, plastic tube 2.1 mm internal diam- eter, by endometrial suction after gonad catheterization (Bromley et al., 2000). These samples were taken from sedated captive Atlantic cod available from a separate study carried out at IMR during March 2004 to deter- mine the rate of transition from normal to advanced stage atretic follicles. Each biopsy sample was fixed as above and examined as an unstained and stained whole mount to compare the intensity of follicular atresia found in both preparations. Intensity of atresia (la) was defined as la = Ni/(Ni +Nj), (2) where Ni and Nj refer to alpha atretic and normal vitel- logenic follicles, respectively. The alpha atresia and more advanced beta follicular stages have been defined previously based on the frag- mentation or absence of the chorion (Witthames and Greer-Walker, 1995;Witthames, 2003) following previ- ous studies (Hunter and Macewicz, 1985b). After scor- ing the intensity of atresia the whole mounts were in- filtrated in resin and then polymerized slowly at -10°C over a period of 2 hours that all the follicles lay at the base of the mold. At least 25 to 30 sections of 5 pm were cut and discarded in order to take a section within 125 to 150 pm from the base of the mold to transect all the follicles present in the sample. This section was stained by the PAS Mallory trichrome method to identify and count the transected follicles. Fecundity maturation and down regulation In order to study the change in fecundity during matu- ration D, and atresia data were taken from Atlantic cod sampled in the Irish and North Seas between Jan- uary and March during 2003 and 2004 (Table 1), and examined in two ways. In the first case the standing stock of atretic follicles (la) was measured as preva- lence (proportion of fish containing alpha atretic fol- licles) and relative intensity (/a/whole body weight g) as described previously (Witthames and Greer-Walker, 1995). The atresia was determined in histological sec- tions stained by PAS Mallory trichrome. Secondly the overall impact of atresia on relative fecundity Fbw = F/total body weight in g) during maturation was deter- mined by assessing the reduction of Fbw in relation to Df as recently described (Thorsen et al., 2006). Fbw - axLn (Df) + b. (3) An additional data set (Table 1) was also available from an annual egg production survey of Atlantic cod, Euro- pean plaice, and common sole biomass (Armstrong et al., 2001) to assess whether down regulation also occurs in other species with a similar fecundity development process. This data set contained details of fecundity, fish length (cm) total, and ovarian weight (g) for each species and was used to calculate Fbw in each case. Df was predicted from Fow using the ovarian weight and fecundity data in Equation 4 (below) adjusted to make Fow the independent variable. Autodiametric calibration A regression line (based on ln-transformed data) was established for each species and institute (Thorsen and Kjesbu, 2001) between Df and Fow using the following formula where a and b are equation constants. Ln Fow = a x In Df + b. (4) In one data set the parameters showed some degree of noncovariance and a second polynomial function (ln Df2) was fitted to the data where a, b and c are constants: Ln Fow = a x \n Df + b x \n Dj + c. (5) Witthames et a I.: Advances in methods for determining fecundity in marine fishes 153 Table 3 Comparison of mean (standard error [SE]) follicle diameter (33^ pm), number of follicles per gram ovary (F0UI), circularity, and fecundity (F * ovary mass [g]), found in samples taken with either a pipette (P, from fresh tissue) or gravimetric method (G, from fixed tissue) from fish caught in the Irish Sea by commercial vessel during 2007. In each case five replicates were taken from a central location of the same ovary in ripe mature Atlantic cod (Gadus morhua, cod m), hydrated Atlantic cod (cod h), ripe mature Atlantic haddock (Melanogrammus aeglefinus, had), and European plaice ( Pleuronectes platessa, pie). The mass (g) of the ovary fresh (ovary F) and after storage in fixative for longer than 14 days (ovary S) is also shown. Parameter Sample method Cod m (n= 5) Cod h (ti = 5) Had (n = 5) Pie (72 = 5) Paired 7-test 77 = 15 P Df P 797(3) 944(5) 571 (4) 1081 (10) 0.0026 G 777 (3) 904(8) 560(1) 1085(7) Fow P 3613 (90) 1460(338) 9355 (338) 1196(38) 6.652-10 G 3975 (149) 1706(33) 10394(149) 1280(17) Circularity P 0.986(0.001) 0.995 (0.001) 0.985 (0.001) 0.990 (0.001) 0.0015 G 0.966(0.004) 0.983(0.001) 0.961 (0.002) 0.977 (0.002) Fecundity P 419,132 388,460 205,804 44,260 G 443,191 421,424 218,664 45.980 Ovary F 116.0 266.0 22.0 37.0 Ovary S 111.5 247.0 21.0 35.9 Fecundity (F) was calculated from the product of ovarian weight ( O ) and Fow. Statistics Regression analysis was carried out using the “R”(vers. 2.5.0, Free Software Foundation, Boston, MA) and residuals were plotted to check there was no systematic pattern suggesting that the models should be further refined. The coefficient of variation (CV) was determined for predictions with new data to examine the precision of the fecundity estimate for a range of follicular sizes typical for each species. Results Ovarian sampling, follicle measurement equipment, and homogeneity In the course of more extensive use at sea, the pipettes performed well taking samples from maturing ovaries providing that they contained vitellogenic follicles vis- ible to the unaided eye (>400 pm). However, ovaries that were close to being spent or immature did not yield quan- titative samples because the connective tissue attached to the follicles pulled the sample out of the pipette as it was withdrawn from the ovary. When this occurred it was clear that the glass tube was only partially filled and the sampling process could be repeated to fill the pipette to avoid under sampling although this was not always successful. In summary we found that replicate subsamples of 25 and 100 pL tissue taken with the pipette from a Atlantic cod ovary equated to a gravi- metric sample of 26.0 mg (CV=1.8%, 72=10) and 106.0 mg (CV=3.7%, 72=10), respectively. Compared to fixing the ovary whole for the gravimet- ric method, the pipette procedure for collecting ovar- ian subsamples was found to significantly increase (P=0.003, P<0.0001, and P=0.002) Dp circularity of fol- licles (Table 3), and decrease Fow, respectively. Ovarian weight after fixation over 63-65 days showed a small decease (95% SE = 0.9) that was not apparently related (P= 0.55, n = 4) to the amount of NBF used to fix the ovary. After accounting for a reduction in ovarian mass the overall reduction in fecundity, determined from the pipette samples, was 5.7% (SE = 0.3) less compared to gravimetric samples taken from the same ovary fixed whole. There was a very significant difference in Fow, Dp and Lc means (P<0.001) between fish but there was no consistent trend either between the pair of ovaries or within the ovary at three sites (anterior, middle, and posterior) where samples were taken (Fig. 1). In two out of the seven fish there was a site effect (left, posterior, and right middle) on Df and Lc, but their rank order was reversed at other locations. It was also noticed that in more mature Atlantic cod ovaries, where the Lc was larger, the CV of Dp and Lc amongst replicates also increased. Similarly, sampling site (anterior, middle, or posterior part of one ovary) in 103 European hake indicated that Fn, , either classified by cortical alveoli, early, or late vitellegenic follicle development stages, or all classes combined, was not related to ovarian position (P=0.133, 0.149, 0.789, 0.101, respectively). Stain evaluation Compared to unstained follicles the use of each stain to color European hake, Atlantic cod, Atlantic mackerel, and European plaice follicles increased the efficiency of image analysis measurement, particularly of semitrans- 154 Fishery Bulletin 107(2) parent objects such as PVF, cortical alveoli, hydrated, and POFs. After PAS staining, manual measurements of follicle diameter (Fdm) compared closely to automatic measured follicle diameter (Fdi) calculated from the image analysis: Fdi = 1.0016 xFdm + 0.0405 (6). (n=42, r=0.997, 220 < manual reading < 1900 pm) Although the eosin solution stained both vitellogenic and hydrated follicles in European plaice, it was much less effective when applied to either Atlantic cod or Atlantic mackerel follicles. A further disadvantage was that the stain was not bound by a chemical reaction and tended to leach out more rapidly compared to PAS stained tissue. This could be countered by extensive washing but this LA LM LP RA RM RP Fish mean Sample site in ovary Figure 1 Comparison of mean follicle diameter (A) and leading cohort diameter (B) ±2 standard error taken from three sites, anterior (LA and RA), middle (LM and RM), and posterior (LP and RP), in each pair (left L or right R, respectively) of seven ovaries of Atlantic cod ( Gadus morhua) caught by commercial vessel using gill nets landing at Andenes, Norway in 2007 (Table 1). Lines join the sites for each fish. Approximately 200 unstained follicles were measured using image analysis by the Institute of Marine Research Norway at each site. Fish mean refers to the overall mean follicle diameter of all sites for each fish. progressively affected the follicle size determined by image analysis. The rose bengal stain was also based on affinity rather than chemically bound and excess stain had to be washed from the sample. It was an effective aid to automatic measurement of PVF and POFs from Atlantic cod, European hake, and Atlantic mackerel, though the coloration was not as intense compared to PAS. Also the PAS stain was particularly useful when applied to whole mounts from Atlantic cod making it easier to identify the outline of small POFs compared to PVF that were less intensely stained. Comparison of whole-mount method with histological method A whole mount prepared from a female Atlantic cod caught during ovulation (Fig. 2A) revealed small POFs from earlier ovulations (Fig. 2B left) and also much larger POFs (Fig. 2B right). The latter appeared as large round membrane structures, about the size of hydrated follicles formed by the thecal and granulosa layers that remain in the ovary to form the POF. A burst zone in the circular membrane was also visible, probably made during the expulsion of the ripe egg. POF shape and morphology was equivalent in whole mounts and section and characterised by deep red staining in section and denser grey scale in stained and unstained whole mount respectively (Fig. 2B left and middle). PVFs in unstained whole mounts appeared quite translucent with a central nucleus that was consistent with their shape and form in section and easily distinguished from POFs. Atretic follicles were rather easily identified in un- stained whole mounts and their morphology could be equated to that seen in histological section (Fig. 3). Comparing these preparations the chorion appeared to be progressively broken down and provided a useful criterion to identify late alpha atretic follicles in whole mounts. A comparison of alpha atretic intensity between the two methods (Fig. 4) indicated that the whole mount preparation could provide an indication of both preva- lence and intensity of atresia. Fecundity maturation and down regulation Atlantic cod fecundity data from the North and Irish Sea collected in 2003 and 2004 (Table 1) was ana- lysed to investigate whether the relative fecundity declined during maturation. As the ovary matured and Df increased from 350 to 800 pm the prevalence of atresia increased (Fig. 5). Relative intensity of atresia was absent in ovaries with a Df of 350 pm and tended to remain at a fairly low level but with one much higher value when the mean follicular diameter was 650 pm. Also during this maturation period there was a drop in predicted mean relative fecundity for all fish in the sample amounting to 49.6% whilst fecundity diameter measured manually or automatically (Fd) increased from 355 to 794 pm. Analysis of the data from the 1995 survey indicated that fecundity was overestimated between 11% (Atlantic cod) and 13% for European plaice Witthames et al.: Advances in methods for determining fecundity in marine fishes 155 A 15 i 400 600 800 1000 1200 1400 1600 Follicle diameter (pm) B Figure 2 Morphology of ovary samples of Atlantic cod ( Gadus morhua) taken from a research vessel trawl catch in the Irish Sea during 2003 shown as whole mounts and histological preparations. (A) The size frequency distribution measured in a whole mount of residual fecundity (black bars) and postovulatory follicles (POFs striped bars). The POFs from the most recent and previous ovulations appear to the right and left respectively of the residual fecundity. (B) (right image) shows the appearance of periodic acid-Schiff stained POF in whole mount from the most recent ovulation with an arrow pointing to the burst zone in the follicle membrane. B (left image) POFs from previous ovulations shown in histological section stained with PAS Mallory (left) and as a whole mount (right). The white and black arrows point to POFs and previtellogenic follicles respectively and box arrows show vitellogenic (V) and early hydration stage (H) follicles. The two scale bars on the left indicate 500 whilst the bar on the right shows 1000 pm. and common sole (Fig. 6, Table 4). From the perspective of fecundity methodology, the measurement of Fd as well as the standing stock of fecundity make it possible to adjust the fecundity to the same point in matura- tion, defined by mean follicle size, close to the start of spawning. Autodiametric calibration The seven species examined, even when in an advanced stage of maturity, contained very different forms of fecundity size frequency distribution (Fig. 7) ranging from normal (Atlantic herring, European plaice, and redfish) to a more skewed shape (European hake and Atlantic mackerel). In two species (Atlantic cod and European plaice) samples with a hydrated, bimodal distribution were also included in the data set. The equations (Table 5) from the regression analy- sis, based on the autodiametric calibration applied individually to Atlantic cod, Atlantic mackerel, Atlan- tic herring, European plaice, and redfish (Fig. 8) for each institute, made it possible to predict Fow (Eqs. 4 and 5) with high precision in most cases (Table 6). European hake and Atlantic mackerel are examples where the vitellogenic follicle distribution is continu- ous extending down to overlap with the PVF popula- tion (Fig. 7) and produced a higher CV to predict Fow from using Equation 2. In the case of European plaice, Atlantic cod, and European hake, some ova- ries contained both maturing and hydrated follicles 156 Fishery Bulletin 107(2) exhibiting a bimodal frequency distribution, but in each case the autodiametric calibration made it possi- ble to make estimates of Fow with an acceptable level of precision. Equation 5 gave a small but significant (PcO.OOOl) improved fit, but only for Atlantic cod with hydrated follicles, and reduced the CV of Fow esti- mates predicted from 450 to 1050 pm D,. The overall precision after inserting an ascending series of Df in Equations 4 and 5, spanning the range found in each species was always better than a CV of 3% based on a prediction for new data. A combination of the data in a general calibration curve (Fig.9A) is provided to show that the auto-dia- metric method has general application and may be used with other species. However when compared against the individual species model there was difference in the pre- dictions by up to 20% both within species and between institutes (Table 6). The variance was greater in the fish with a continuous follicular distribution, especially in the case of European hake (Fig. 9, B and C). Discussion Our results show that the pipette method for sampling fresh ovaries at sea can be used to replace the need to return the whole ovary for the gravimetric fecundity method (Bagenal and Braum, 1968), provided ovarian weight can be recorded precisely onboard. Although the pipette fecundity was slightly lower (94.7%) compared to the gravimetric fecundity, we feel that the scale of difference can be easily nullified by a small correction factor and is small compared to the reported variability in fecundity over time (Rijnsdorp, 1991) and space (Witthames et al., 1995). Our confidence in making this statement is increased because of a direct comparison between both methods for the same ovary and because the autodiametric calibrations are very similar without large residuals attached to either method. A previ- ous report described a cut down plastic syringe to suck up standard sized ovarian samples of 1.54 g (CV=3.7 rc=155), but the commercially available alternative described in this paper has two advantages: 1) it is already calibrated for a range of sample sizes (25, 50, 100, and 200 pL), and 2) it is suited to taking small samples appropriate for fecundity determina- tion in species such as Atlantic mackerel and European hake. In our results ovarian weight showed on average a small decline (-5%) from fresh to fixed weight for each species which was considerably different from a previous report (Klibansky and Juanes, 2007) at +5% or less. The reasons for the difference are not apparent but do not involve the ratio of fixative to weight of ovarian tissue because the range used in this work (1.7 to 9.1 times NBF to ovar- ian weight) spans the ratio of four times where a positive weight change was recorded. Collection of fecundity samples in this way has clear advantages: 1) require small amounts (1.2 compared to more than 5000 ml for species like Atlantic cod) of NBF (classed as a carcinogen), 2) reduced exposure because of the smaller free surface for evaporation, 3) lower environmental impact for disposal of fixed tissue and waste fixative, 4) it is more feasible to collect fecundity samples on com- Figure 3 Appearances of atretic oocytes taken from 2 year old aquaculture reared Atlantic cod ( Gadus morhua) at Institute IMR in 2004. (A) Image taken from an unstained whole mount prepared from an ovary biopsy containing high levels of alpha atretic follicles (dashed circles), beta atretic follicles (dotted circles), and normal vitellogenic follicles (black circles). (B) Histological section of the biopsy in A showing the same classes of follicle (outlined using the same line key as A) after processing into histological section. In each case arrows point to the disintegrating chorion used for classification of alpha atretic follicles. The scale bars (top left of A and B) = 500 pm. Witthames et al.: Advances in methods for determining fecundity in marine fishes 157 45 0 +• 1 1 1 0 15 30 45 Histology (%) Figure 4 Comparison of atretic follicles as a percentage of atretic follicles divided by the sum of vitellogenic and atretic follicles in whole mount and histology, respectively found in 2 year old aquaculture reared Atlantic cod (Gadus morhua) at Institute IMR in 2004. The equation for the fitted line is W=Hx0.85+4.20 (n = 18, r2=0.79, P <0.001). mercial vessels, 4) easily portable samples that can be distributed at lower cost to facilitate exchange of samples between laboratories carrying out international egg production based stock assessments (Armstrong et al., 2001) and, not least, 5) better preservation of his- tological detail to identify and determine proportions of atretic and postovulatory follicles. Our finding that POF in recently ovulated ovaries are spherical and not collapsed, compared to previous studies, may be attributed to less compression in a biopsy compared to ovaries fixed whole. A small difficulty in not returning the whole ovary back to the laboratory for subsampling means the ovary must be weighed at sea, sometimes in rough conditions, but this can be achieved using motion-compensated bal- ances. Although this equipment is perhaps beyond some research budgets it has been successful even on com- mercial vessel in rough conditions providing the mass for weighing is less than 75% of the upper weighing range. Also balances are available with a resolution of 0.01 g making it feasible to weigh ovaries from probably all commercial species with an acceptable accuracy. If it is not feasible to use motion-compensated balances, it has been reported that ovaries can be returned for weighing after the end on of the cruise providing the ovary is packed to prevent water loss (Klibansky and Juanes, 2008). A further recommendation, in connection with using the pipette, is to complete at least duplicate samples and the follicles should not be larger than the pipette internal bore of 2 mm although exception can be made for hydrated follicles just larger than 2 mm. In this study it was shown that the ovary is homog- enous in regard to Fow and Df both for Atlantic cod and 03 c n o 03 0) > 0) cc 60 i B 50 - 40 - 30 - 20 - 10 - 0 - o o o 8 o 1600 >, 1400 - ? 1200 • 3 iooo - ® 800 ■ d) ■g 600 • I 400 ■ 00 200 - 0 ■ 300 400 500 600 700 800 Mean follicle diameter (pm) Figure 5 Down regulation of fecundity during mat- uration in North and Irish Sea Atlantic cod ( Gadus morhua) collected in 2003 and 2004. (A) Prevalence of atretic follicles (proportion of fish with atresia) plotted against the mean follicle diameter of the fecundity (Dp pm). (B) Relative intensity of atretic follicles (atretic follicles per g female whole weight) plotted against Dp, (pm). (C) The decline in relative fecun- dity (F6u,=fecundity/body weight [g] ) in relation to Dp (pm). The fitted line is from the equation xlnZT+ b where a=-581.9 and b = 4360 (r2 = 0.30, P for a and b <0.0001). Open circles are used in A and B because they refer to regressing follicles and filled circles are used in C to denote normal vitellogenic follicles. European hake. Homogenous packing has also been reported for hydrated oocytes prior to ovulation in Eu- ropean hake (Murua et al., 2006). The posterior region of the ovary is the most variable and in some fish this part can be packed with significantly different sized follicles and should be avoided. However, it should not be assumed that Fd is universally independent of loca- tion because small differences (2%) in Fd heterogeneity have been reported in flatfish species such as yellowfin 158 Fishery Bulletin 107(2) sole Limanda asper (Nichol and Acuna, 2001) and Eu- ropean plaice (Kennedy et al., 2007). Samples used for the auto- diametric calibration and in subsequent deter- mination of Fow should have the same fixation history because fixing conditions affect Dp Fow, and circularity and also affects the ovarian weight (Klibansky and Juanes, 2007) used to raise Fow to fecundity. We would not recommend the general use of PAS stain for image analysis in mature fish because it obscures < 300 400 500 600 700 800 o> 600 800 1000 1200 300 400 500 600 700 800 900 Mean follicle diameter (pm) Figure 6 Reduction in relative fecundity (Fbw) during maturation, measured as mean follicle diam- eter (Dp pm) in Atlantic cod ( Gadus morhua ), European plaice (Pleuronectes platessa), and common sole (Solea solea) (A-C, respectively) collected from the Irish Sea during 1995. The equation for the fitted line and regression coef- ficients are shown in Table 4. the chorion detail which is used to classify atretic from normal vitellogenic follicles (Kjesbu et al., 1991). The main advantages of PAS, compared to the other stains evaluated, was that it was the most color fast, worked with all the species where it was tried, and provided specific staining to color more transparent objects such as cortical alveoli, hydrated, and postovulatory follicles. It is however more laborious to apply, but has been suc- cessful in all applications where it has been tried previ- ously (Kennedy et al., 2007) and performed well in the comparison of manual versus automatic measurements. Similar results have also been found for nonstained fol- licles, although not reported in the results section. Based on our results and earlier reports (Witthames and Greer-Walker, 1995; Kurita et al., 2003; Thorsen et al., 2006; Kennedy et al., 2007) fecundity is down regulated by the production of atretic follicles during maturation. If samples are taken close to spawning season, down regulation is not significant (Oskarsson and Taggart, 2006), but the timing of sampling should be considered especially when studying multiyear col- lections for example: Atlantic cod (McIntyre and Hutch- ings, 2003), European plaice (Horwood et al., 1986; Rijnsdorp, 1991) and common sole (Witthames et al., 1995). Using the autodiametric method, it was pos- sible to predict Dp providing data on ovarian weight and fecundity is reported using a rearranged Equa- tion 4. This method was used in this study for 1995 survey data to standardize fecundity for maturity and indicated that the spawning stock biomass of Atlantic cod, European plaice, and common sole may have been overestimated by about 12% based on follicle diameters of 650, 1100, and 600 pm, respectively. Although we consider that follicular atresia was an important cause of negative fecundity residuals in this study, we do not exclude an alternative explanation that more fecund individuals within a fecundity sample produce smaller eggs, and vice versa (i.e., a trade off between fecundity and egg size). Such a trade off is likely in a comparison between stocks such as Atlantic herring (Winters et al., 1993) but has not, to our knowledge, been proven to oc- cur within one stock. One report referring to Atlantic cod from the Norwegian coast (Kjesbu et al., 1996a) indicates that much of the variability in egg size occurs during the final maturation rather than variability in follicular size when final maturation occurs. Overall our view is that the relationship used for fecundity standardisation should be documented, including the follicle size reference point along with the unadjusted results. Different image analysis configurations used by four institutes to collect the autodiametric calibration data produced a low CV of fecundity estimates for new pre- dictions. The data can be accumulated without inter- vention (Thorsen and Kjesbu, 2001) in automatic mode and has utility for a number of species. Since it is an automatic process it is important that all follicular classes of interest are measured with equal selectivity, including cortical alveoli, atretic, or hydrated follicles. We suspect that the cause of the higher fecundity CV Witthames et al. : Advances in methods for determining fecundity in marine fishes 159 40 30 20 European hake: Spawning Atlantic cod: Spawning 1 Atlantic mackerel: Late maturation European plaice: Spawning lunnq 400 800 1200 1600 2000 400 800 1200 1600 2000 400 800 1200 1600 2000 Follicle diameter (pm) 400 800 1200 1600 2000 Figure 7 Follicle number per g of ovary (percentage of total count) per 25-pm class interval follicle diameter found in Atlantic cod ( Gadus morhua), European hake ( Merluccius merluccius ), Atlantic herring ( Clupea harengus), Atlantic mackerel ( Scomber scombrus), European plaice (Pleuronectes platessa), redfish (deep water redfish [ Sebastes mentella ] or golden redfish [Sebastes marinus]) used to produce the autodiametric calibrations (Table 5) by Institutes AZTI (A, stained rose bengal), Cefas (B, stained with periodic acid Schiff’s, except European plaice with eosin), CSIC (C, hake rose bengal, redfish unstained), and IMR (D, unstained). Table 4 Regression coefficients used to fit relative potential fecundity (Fbw) with mean follicle diameter (Dp pm) in the following equation Fbw = a xLn (Dp) + b using data collected from Atlantic cod (Gadus morhua), European plaice (Pleuronectes platessa), and common sole (Solea solea) caught in the Irish Sea during 1995. Species Coefficient SE t P Atlantic cod, n = 54 b 4356 2039 4.6 <0.0001 a -521 317 -1.6 0.1068 European plaice, n=220 b 1578 318 55.0 <0.0001 a -190 46 -4.1 <0.0001 Common sole, n=129 b 3257 929 6.8 <0.0001 a -406 143 -2.8 0.0055 160 Fishery Bulletin 107(2) Table 5 Details of the autodiametric calibration relating follicle number (Fow) to follicle diameter {Dfpx n) using a linear equation lnFoiu= ax In Df+ b and polynomial equation (5) In Fow = axln Df+ bxln Dj +c fitted to data collected by each institute: AZTI, Cefas, CSIC, and IMR for each species: Atlantic cod (Gadus morhua), European hake ( Merluccius merluccius ), Atlantic herring (Clupea harengus), Atlantic mackerel ( Scomber scombrus), European plaice ( Pleuronectes platessa), and redfish (deep water redfish [Sebastes mentella ] or golden redfish [Sebastes marinus]. Species Institute a b c N r2 Atlantic cod Cefas -3.106 28.777 28 0.972 Atlantic cod Cefas polynomial 17.234 -1.5497 -37.864 28 0.986 Atlantic cod IMR -2.700 26.088 47 0.988 European hake AZTI -2.157 22.293 157 0.774 European hake CSIC -2.196 22.758 245 0.780 Atlantic herring IMR -2.718 26.287 23 0.971 Atlantic mackerel Cefas -2.528 25.030 78 0.761 European plaice Cefas -2.910 27.442 150 0.980 Redfish CSIC -2.551 25.040 147 0.948 General (excluding European hake) All institutes except AZTI -2.750 26.371 475 0.979 reported for species like European hake or Atlantic mackerel maybe attributed to the ovary being packed with a larger, and perhaps more variable partial vol- ume of PVF associated with a continuous follicular distribution. Further analysis to determine the source of variation in the autodiametric calibration for fish with a continuous follicular frequency distribution is therefore considered worthwhile. Calibration data that included spawning Atlantic cod was best described by a polynomial model, although the additional term was not significant for the European plaice, even though the data included fish with hydrated follicles and POF. The difference may arise because European plaice produce fewer egg batches, about five (Urban, 1991), compared to between 14 and 21 in Atlantic cod (Kjesbu et al., 1996b). Thus, residual POFs in Atlantic cod ovaries should take up increasingly more space in the ovary towards the end of the spawning season changing the relative partial volume taken up by residual vitellogenic follicles. An alternative to full automation is to use a semi- automatic analysis so that follicles that are not meas- ured by the automatic analysis can still be measured manually. In practice, the dominant fecundity follicles were measured in automatic mode and then other follicular types, such as POFs or atretic follicles, are manually assigned and measured accumulating the measurements in user defined classes. This informa- tion can be used for more qualitative aspects, such as an overview of atresia intensity or confirming fish are at an advanced state of maturity, and also to provide a means to exclude fish that have started spawning. Our experience shows that POFs will arise from a synchronous ovulation that will produce a cohort of POFs of similar size and shape thus mak- ing their identification more certain. In practice we keep a tally of identified POFs in a separate class and reject the fish from the fecundity data set to apply the annual egg production method if five or more POFs with similar structures are found. The hydrated cohort were split from the vitellogenic mode to determine the batch fecundity by inspection of the frequency distribution produced from the follicular measurements. This provides a further advantage for the study of batch fecundity because it is easier to see and separate the next batch compared to the traditional gravimetric method described previously (Hunter and Macewicz, 1985a). In conclusion the present study has shown that im- age analysis and the autodiametric method have wid- er application than originally reported (Thorsen and Kjesbu, 2001; Klibansky and Juanes, 2008). Although one report (Friedland et al., 2005) indicated caution in this respect, the range of spawning strategies and institutes participating in this study indicate that for species with a discontinuous follicular frequency distribution, the method is also reliable. However, the authors have demonstrated that a calibration should be done to validate the method in all new applications whether it involves new species, equipment, or situa- tion. The use of the pipette makes it possible to take quantitative fecundity samples in situations were ac- curate balances, measuring to an accuracy of 0.1 mg, will not function. In addition this provided a means to calibrate the autodiametric method for routine qual- ity control and substantially reducing the use of toxic fixative. Substantial histology costs can be avoided by improving the interpretation of whole mounts and the approach has great utility to study the fate of fecun- dity during the spawning season. Witthames et al.: Advances in methods for determining fecundity in marine fishes 161 100000 Deep water redfish and golden redfish (C) Atlantic herring (D) 10000 1000 100 ' ' ' ' ' ' ' • ■ ' * ' > 200 600 1000 1400 1600 600 1000 1400 1600 Mean follicle diameter (pm) Figure 8 Autodiametric calibrations shown as scatter plots and fitted lines for Atlantic cod ( Gadus morhua), European hake ( Merluccius merluccius), Atlantic herring < Clupea harengus), Atlantic mackerel ( Scomber scombrus ), European plaice ( Pleuronectes platessa), redfish (deep water redfish [Sebastes mentella] or golden redfish [Sebastes marinus ]) used to produce the autodiametric calibrations (Table 5) by Institutes AZTI (A stained Rose Bengal), Cefas (B stained with periodic acid Schiff’s, except European plaice with eosin), CSIC (C hake Rose Bengal, redfish unstained), and IMR (D unstained). Acknowledgments This study was jointly funded under European Union Frame Work V Q5RS-2002-01825 and the Institutes in England (Department of the Environ- ment, Food, and Rural Affairs), Norway (Institute of Marine Research), and Spain (Consejo Superior de Investigaciones Cientificas, and AZTI Tecnalia (publication number 424)). The first author would like to acknowledge the contribution of J. Pilking- ton who wrote the software code for Myrmica. This software will be distributed at no charge under a Free Software Foundation License at www.myrmica. co.uk. 162 Fishery Bulletin 107(2) cc ~0 c n CD 100 -| 75 - -75 B -100 i r 100 -| -100 "I i i 1 i 1 1 1 1 1 200 400 600 800 1000 1200 1400 1600 1800 2000 Follicle dimeter (pm) Figure 9 (A) General autodiametric calibration for Atlantic cod ( Gadus morhua), Atlantic herring ( Clupea harengus), Atlantic mackerel ( Scomber scombrus), European plaice (Pleuronectes platessa ), redfish (deep water redfish [Sebastes mentella ]) or golden redfish ( Sebastes mari- nus) used to produce the autodiametric calibrations (Table 5) by Institutes Cefas (B, stained with periodic acid Schiff’s, except European plaice with eosin), CSIC (C, hake rose bengal, redfish unstained), and IMR (D, unstained) based on the combined data Df and follicle number (Fow). The parameters (Table 5) for the fitted line were based on Equation 4 (In Fow=ax In Df +b). P for a and b was <0.001. (B) Plot of normalised residual (F-OJ/F x 100 where Os=observed value for Atlantic cod, Atlantic herring, Atlantic mackerel, European plaice, and redfish (deep water redfish ( Sebastes mentella ) or ocean perch (Sebastes marinus) and F=fitted value based on Faw against (Eq. 4). (C) Plot of normalized residual ( F—Oh)/F x 100 where 0/; = observed fecundity for European hake ( Merluccius merluccius), and F=fitted fecundity value based on the observed Fd for hake sub- stituted in the general model (Eq. 4). Witthames et at: Advances in methods for determining fecundity in marine fishes 163 Literature cited Andersen, T. E. 2003. Unbiased stereological estimation of cell numbers and volume fraction: the disector and the principles of point counting. In Fisken Havet 2003(12): Report of the working group on modern approaches to assess fecundity and maturity of warm- and cold-water fish and squids (O. S. Kjesbu, J. R. Hunter, and P. R. Witthames, eds.), p. 11-18. The Institute of Marine Research, Bergen, Norway. Armstrong, M. J. P., P. Conolly, R. D. M. Nash, M. G. Pawson, E. Alesworth, P. J. Coulahan, M. Dickey-Collas, S. P. Milligan, M. F. O’Neill, P. R. Witthames, and L. Woolner. 2001. An application of the annual egg production method to estimate the spawning biomass of cod (Gadiis morhua L), plaice ( Pleuronectes platessa L) and sole ( Solea solea L.) in the Irish Sea. ICES J. Mar. Sci. 58:183-203. Bagenal, T. B., and E. Braum. 1968. Eggs and early life history. In Methods of assess- ment of fish production in fresh waters (W. E. Ricker, ed.), p. 165-201. Blackwell Sci. Publ., Oxford, U.K. Beverton, R. J. H., and S. J. Holt. 1957. On the dynamics of exploited fish populations. Fish. Invest Ser. II Mar. Fish. G.B. Minist. Agric. Fish Food 19:1-553. Bromley, P. J., C. Ravier, and P. R. Witthames. 2000. The influence of feeding regime on sexual matu- ration, fecundity and atresia in first time spawning turbot. J. Fish Biol. 56:264-278. Emerson, L. S., M. Greer -Walker, and P. R. Witthames. 1990. A stereological method for estimating fish fecundity. J. Fish Biol. 36:721-730. Friedland, K. D., D. Ama-abasi, M. Manning, L. Clarke, G. Kligys, and R. C. Chambers. 2005. Automated egg counting and sizing from scanned images: rapid sample processing and large data volumes for fecundity estimates. J. Sea Res. 54:307-316. Greer -Walker, M., P. R. Witthames, and I. Bautista de los Santos. 1994. Is the fecundity of the Atlantic mackerel ( Scomber scombrus: Scombridae) determinate? Sarsia 79:13— 26. Horwood, J. W., R. C. A. Bannister, and G. J. Howlett. 1986. Comparative fecundity of North Sea plaice ( Pleu- ronectes platessa L.). Proc. R. Soc. Lond. Ser. B. 228:401-431. Hunter, J. R., and B. J. Macewicz. 1985a. Measurement of spawning frequency in multiple spawning fishes. In An egg production method for esti- mating spawning biomass of pelagic fish: application to the northern anchovy, Engraulis mordax (Lasker, ed.), p. 79-94. NOAA Tech. Rep. NMFS 36, Spring- field, VA. 1985b. Rates of atresia in the ovary of captive and wild northern anchovy, Engraulis mordax. Fish. Bull. 83:119-136. 2003. Improving the accuracy and precision of reproduc- tive information used in fisheries. In Fisken Havet 2003(12): Report of the working group on modern approaches to assess fecundity and maturity of warm- and cold-water fish and squids (O. S. Kjesbu, J. R. Hunter, and P. R. Witthames, eds.), p. 57-68. The Institute of Marine Research, Bergen, Norway. Hunter, J. R., B. J. Macewicz. N. C. H. Lo, and C. A. Kimbrell. 1992. Fecundity, spawning, and maturity of female Dover sole Microstomus pacificus, with an evaluation of assump- tions and precision. Fish. Bull. 90:101-128. Kennedy, J., P. R. Witthames, and R. D. M. Nash. 2007. The concept of fecundity regulation in plaice ( Pleu- ronectes platessa L.) tested on three Irish Sea spawning populations. Can. J. Fish. Aquat. Sci. 64:587-601. Khoo, H. K. 1979. The histochemistry and endocrine control of vitel- logenesis in goldfish ovaries. Can. J. Zool. 57:617- 626. Kjesbu, O. S. 1991. A simple method for determining the maturity stages of Northeast Arctic cod ( Gadus morhua L.) by in vitro examination of oocytes. Sarsia 75:335—338. Kjesbu, O. S., J. Klungsoyr, H. Kryvi, P. R. Witthames, P. R., and M. Greer-Walker. 1991. Fecundity, atresia, and egg size of captive Atlan- tic cod ( Gadus morhua) in relation to proximate body composition. Can. J. Fish. Aquat. Sci. 48:2333- 2343. Kjesbu, O.S., H. Kryvi, and B. Norberg. 1996a. Oocyte size and structure in relation to blood plasma steroid hormones in individually monitored, spawning Atlantic cod. J. Fish Biol. 49:1197-1215. Kjesbu, O. S., P. Solemdal, P. Bratland, and M. Fonn. 1996b. Variation in annual egg production in individual captive Atlantic cod ( Gadus morhua). Can. J. Fish. Aquat. Sci. 53:610-620. Kjesbu, O. S., and P. R. Witthames. 2007. Evolutionary pressure on reproductive strategies in flatfish and groundfish: Relevant concepts and meth- odological advancements. J. Sea Res. 58:23-34. Klibansky, N., and F. Juanes. 2007. Species-specific effects of four preservative treatments on oocytes and ovarian material of Atlan- tic cod (Gadus morhua), haddock ( Melanogrammus aeglefinus), and American plaice ( Hippoglossoides platessoides). Fish. Bull. 105:538-547. 2008. Procedures for efficiently producing high-quality fecundity data on a small budget. Fish. Res. 89:84- 89. Kurita, Y., O. S. Kjesbu, and S. Meier. 2003. Oocyte growth and fecundity regulation by atresia of Atlantic herring ( Clupea harengus) in relation to body condition throughout the maturation cycle. J. Sea Res. 49:203-219. Lo, N. C. H., J. R. Hunter, H. G.. Moser, P. E. Smith, and R. D. Methot. 1992. The daily fecundity reduction method: A new pro- cedure for estimating adult fish biomass. ICES J. Mar. Sci. 49:209-215. Lockwood, S. J., J. H. Nichols, and W. A. Dawson. 1981. The estimation of a mackerel ( Scomber scomber L.) spawning stock size by plankton survey. J. Plankton Res. 3:217-233. Marshall, T. C., O. S. Kjesbu, N. A. Yaragina, P. Solemdal, and 0. Ulltang. 1998. Is spawning stock biomass a sensitive measure of the reproductive potential of Northeast Arctic cod? Can. J. Fish. Aquat. Sci. 55:1766-1783. Marshall, T. C., N. A. Yaragina, Y. Lambert, and O. S. Kjesbu. 1999. Total lipid energy as a proxy for total egg produc- tion by fish stocks. Nature 402:288-290. McBride, R. S., and P. E. Thurman. 2003. Reproductive biology of Hemiramphus brasiliensis 164 Fishery Bulletin 107(2) and H. balao (Hemiramphidae): Maturation, spawning frequency, and fecundity. Biol. Bull. 204:57-67. McIntyre, T. M., and J. A. Hutchings. 2003. Small-scale temporal and spatial variation in Atlantic cod ( Gadus morhua ) life history. Can. J. Fish. Aquat. Sci. 60:1111-1121. Murua, H., G. Kraus, F. Saborido-Rey, P. R. Witthames, A. Thorsen, and S. Junquera. 2003. Procedure to estimate fecundity of marine fish species in relation to their reproductive strategy. J. Northwest Atl. Fish. Sci. 33:33-54. Murua H., P. Lucio, M. Santurtun, and L. Motos. 2006. Seasonal variation in egg production and batch fecundity of European hake Merluccius merluc- cius (L.) in the Bay of Biscay. J. Fish Biol. 69:1304- 1316. Murua, H., and L. Motos. 2006. Reproductive strategy and spawning activity of the European hake, Merluccius merluccius L., in the Bay of Biscay. J. Fish Biol. 69:1288-1303. Murua, H., and F. Saborido-Rey. 2003. Female reproductive strategies of marine fish species of the North Atlantic. J. Northwest Atl. Fish. Sci. 33:23-31. Nichol, D. G., and E. I. Acuna. 2001. Annual and batch fecundities of yellowfin sole, Limanda aspera, in the eastern Bering Sea. Fish. Bull. 99:108-122. Oskarsson, G. J., O. S. Kjesbu, and A. Slotte. 2002. Predictions of realized fecundity and spawning time in Norwegian spring-spawning herring ( Clupea harengus). J. Sea Res. 48:59-79. Oskarsson, G. J., and C. T. Taggart. 2006. Fecundity variation in Icelandic summer- spawning herring and implications for reproductive potential. ICES J. Mar. Sci. 63:493—503. Parker, K. 1980. A direct method for estimating northern anchovy, Engraulis mordex, spawning biomass. Fish. Bull. 78:541-544. Ramsay, K., and P. R.Witthames. 1996. Using oocyte size to assess seasonal ovarian development in Solea solea (L.). J. Sea Res. 36:275— 283. Rijnsdorp, A. D. 1991. Changes in fecundity of female North Sea plaice (Pleuronectes platessa L.) between three periods since 1900. ICES J. Mar. Sci. 48:253-280. Simpson, A. C. 1951. The fecundity of the plaice. Fish. Invest. Ser. II Mar Fish. G.B. Minist. Agric. Fish Food 17:1-27. Thorsen, A., and O. S. Kjesbu. 2001. A rapid method for the estimation of oocyte size and potential fecundity in Atlantic cod using computer-aided particle analysis system. J. Sea Res. 46:295-308. Thorsen, A., C. T. Marshall, and O. S. Kjesbu. 2006. Comparison of various potential fecundity models for north-east Arctic cod Gadus morhua, L. using oocyte diameter as a standardizing factor. J. Fish Biol. 69:1709-1730. Tyler, C. R., and J. P. Sumpter. 1996. Oocyte growth and development in teleosts I Review], Rev. Fish Biol. Fish. 6:287-318. Urban, J. 1991. Reproductive strategies of North Sea plaice, Pleu- ronectes platessa, and North Sea sole, Solea solea : batch spawning cycle and batch fecundity. Meeresforschung 33:330-339. Winters, G. H., J. P. Wheeler, and D. Stansbury. 1993. Variability in the reproductive output of spring- spawning herring in the north-west Atlantic. ICES J. Mar. Sci. 50:15-25. Witthames, P. R. 2003. Methods to assess maturity and realized fecundity illustrated by studies on Dover sole Solea solea. In Fisken Havet 2003(12): Report of the working group on modern approaches to assess fecundity and maturity of warm- and cold-water fish and squids (O. S. Kjesbu, J. R. Hunter, and P. R. Witthames, eds.), p. 125-137. The Institute of Marine Research, Bergen, Norway. Witthames, P. R., and M. Greer-Walker. 1987. An automated method for counting and sizing fish eggs. J. Fish Biol. 30:225-235. 1995. Determination of fecundity and oocyte atresia in sole ( Solea solea) (Pisces) from the Channel, the North Sea and the Irish Sea. Aquat. Living Resour. 8:91-109. Witthames, P. R., M. Greer-Walker, M. T. Dinis, and C. L. Whiting. 1995. The geographical variation in the potential fecun- dity of Dover sole Solea solea (L) from European Shelf edge waters during 1991. J. Sea Res. 34:45-58. Witthames, P. R., and C. T. Marshall. 2008. The importance of reproductive dynamics in fish stock assessments. In Advances in fisheries science: 50 years on from Beverton and Holt (A. Payne, J. Cotter, and T. Potter, eds.), p. 306-324. Blackwell Publ., Oxford, U.K. 165 Abstract — The tidal freshwater of Virginia supports anadromous her- ring ( Alosa spp.) spawning runs in the spring; however, their importance as nutrient delivery vectors to the freshwater fish food web remains unknown. The stable isotope sig- natures of fishes from 21 species and four different guilds (predators, carnivores, generalists, and plankti- vores) were examined in this study to test the hypothesis that marine derived nutrients (MDNs) brought by anadromous fish would be traced into the guilds that incorporated them. Spawning anadromous fish were 13C and 34S-enriched (613C and 634S of approximately 18%c and 17.7%», respec- tively) relative to resident freshwater fish. Of the guilds examined, only predators showed 13C and 34S-enrich- ment similar to the anadromous fish; however, some generalist catfish also showed enriched signatures. Specific fatty acid 613C signatures for gizzard shad (Dorosoma cepedianum), blue catfish ( Ictalurus furcatus), and ale- wife (Alosa pseudoharengus), show a 10 %c range among fishes, clearly reflecting isotopically distinct dietary sources. The 613C and <534S distribu- tion and range among the freshwater fishes suggest that both autochtho- nous and allochthonous (terrestrial C3 photosynthetic production and MDN) nutrient sources are important to the tidal freshwater fish community. Manuscript submitted 25 June 2008. Manuscript accepted 20 October 2008. Fish. Bull. 107:165-174(2009). The views and opinions expressed or implied in this article are those of the author and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Anadromous fish as marine nutrient vectors Stephen E. MacAvoy (contact author)1 Greg C. Garman2 Stephen A. Macko3 4 Email address for contact author: macavoy@american.edu 1 Department of Environmental Science American University 4400 Massachusetts Ave. NW Washington, DC 20016 2 Center for Environmental Studies Virginia Commonwealth University 1000 W. Cary St., Suite 111 Richmond, Virginia 23284 3 Environmental Science Department University of Virginia 291 McCormick Rd. Charlottesville, Virginia 22903 4 Program in Geobiology and Low Temperature Geochemistry U.S. National Science Foundation 4201 Wilson Boulevard Arlington, Virginia 22230 Streams in which anadromous fish spawn are often nutrient poor and the spawning anadromous fish may be an important source of nutrients to them (Kline et al., 1993; Wipfli et al., 2003). Sometimes spawning anadromous fish even fertilize near- stream terrestrial environments (Ben-David et al., 1998; Koyama et al., 2005). The spawning fish are frequently semelparous and deliver marine derived nutrients (MDN) to the freshwater as moribund biomass, excreted ammonium ion (NH4+), or through gamete release (Cederholm et al., 1989; Browder and Garman, 1994; Wipfli et al., 2003). Several studies in Alaska and the Pacific Northwest of North America have demonstrated the importance of marine nutrients brought to freshwater streams by anadromous salmonids (Bilby et al., 2003; Kline et al., 1993; Francis et al., 2006). In the Gulf of Mexico, migrating Gulf menhaden (Brevoor- tia patronus) transported estuarine nutrients into inshore environments (Deegan, 1993), and returning salmon contributed to the productivity of Lake Ontario tributaries (Rand et al., 2002). However, less work has been done on the East Coast of the United States where coastal development has been much more intense and the dominant anadromous species ( Alosa spp.; herring (A. aestivalis), American shad (A. sapidissima), and alewife (A. pseudoharengus )) are often not highly abundant (Deegan, 1993; Garman and Macko, 1998). Although the Alosa spp. on the east coast tend towards an iteroparous life cycle rather than a semelparous one, they do experi- ence heavy postspawning mortality (alewife postspawning mortality has been measured as 41% (Havey, 1961) and between 39% and 57% (Durbin et al., 1979)). Because tidal freshwater streams receive nutrients from marine and freshwater primary productiv- ity at different times, the incorpora- tion of these nutrients by consumers may be different depending on feeding guilds. Fish found in the same area in a stream may derive nutrition from local or translocated productivity. In nutrient poor systems, such as East Coast United States tidal freshwater areas, it is important to understand nutrient sources to different feeding guilds (e.g., predators, carnivores, generalists, and planktivores). For more than 20 years now, car- bon and nitrogen stable isotopes (re- 166 Fishery Bulletin 107(2) ported as a ratio of heavy to light isotopes and given d notation with units of %o, see Materials and methods section for more detail) have been used to determine the importance of MDN in freshwater systems, and to characterize the trophic structure within those systems (Kline, et al., 1993; Vander-Zanden et al., 1999). For example, carbon and nitrogen isotopes have shown that anadromous Pacific salmon ( Oncorhynchus spp.) were a significant source of allochthonous nitrogen to coastal streams where spawning occurs (Kline et al., 1993). Hesslein et al. (1991) used sulfur isotopes to differen- tiate freshwater migratory and non-migratory fishes in the Mackenzie River Basin, Canada. On the East Coast of the United States, anadromous river herring ( Alosa spp.) retain their marine isotope signal after spending part of the spring spawning in freshwater, and that some freshwater piscivores are 34S and 13C-enriched after preferentially consuming migrating Alosa spp. during the spawning run (Garman and Macko, 1998; MacAvoy et al., 2000). An additional tool for determining origins and trans- formations of organic material from different sources is the stable isotope ratio of specific compounds. Isolat- ing a specific compound, or class of compounds, then measuring the isotope ratio on those compounds, may offer a more robust technique to trace biologically significant compounds (such as fatty or amino acids) than would be possible from bulk isotope analysis alone. For example, examining the carbon isotopic composition of fatty acids from an animal, particularly essential fatty acids, allows the direct determination of dietary sources that contribute to the fatty acid pool of that animal (Stott et al., 1997). Although bulk isotope analysis can be an effective nutrient tracer in systems with isotopically distinct nutrient sources (Peterson et al., 1985), the isotopes of specific fatty acids may provide more confidence in identifying sources (Canuel et al., 1997). Carnivorous heterotrophs are unable to synthesize fatty acids longer than 18-carbons, nor can they de- saturate carbon-carbon bonds between the ninth and terminal methyl carbon, therefore, these essential fatty acids must be obtained from diet (Olsen 1999). Because essential fatty acids are not influenced by subsequent metabolism within a eukaryotic heterotroph, they re- tain their original isotope composition (Stott et al., 1997). Fatty acids synthesized by marine plankton and incorporated into marine fish would be highly en- riched in 13C relative to those produced by freshwater primary producers or C3 photosynthesis. Addition- ally, short chain fatty acids, used as precursors in the biosynthesis of unsaturated or longer chain saturated fatty acids, should be 13C enriched in relation to bio- synthesized fatty acid products (Murphy and Abrajano, 1994). In this study, the fatty acid nomenclature used is carbon numberinumber of double bonds. For ex- ample, 18:2 is an 18-carbon fatty acid with two points of unsaturation. The desaturation of 16:0 to 16:1 and 18:0 to 18:1-18:2 occurs by a systematic fractionation of roughly 2 %c per desaturation (DeNiro and Epstein, 1977; Monson and Hayes, 1982). Also, studies have shown that the elongation of fatty acids by de novo synthesis results in a 2 %c per 2-carbon acetyl group addition. These fractionations allowed the identifica- tion of fatty acids that were directly incorporated from symbiotic bacterial sources in mussels as opposed to those obtained through de novo synthesis (Murphy and Abrajano, 1994). In this study we compared the d15N, dL3C, 634S of bulk tissues, plus the 613C of specific fatty acids among four guilds of fish plus anadromous Alosa spp. in a tidal freshwater stream on the East Coast of the United States. Our objective was to determine if anadromous fish, captured more than 40 km from the salt-wedge, were isotopically distinct from freshwater residents, and to determine if freshwater guilds showed the incorpora- tion of marine allochthonous organic material. Materials and methods Field collections by boat electrofisher were made in the tributaries and main-stem of the Rappahannock River, VA (within a 40 -mile area between Fredericksburg and Tappahannock, VA) during March and May 1997 and 1998 (Fig. 1). The Rappahannock River is tidal in this region (tidal range: 0.1 to 1 meter) and shares many physicochemical characteristics with other tidal fresh- water rivers in the region (Garman and Nielsen, 1992). Fishes were collected and placed on ice in the field, transported back to the laboratory, and muscle tissue samples were taken, which were then dried for later analysis. Analysis of the sulfur and compound specific fatty acid samples took several years and were completed by 2002. The fishes were placed into four different guilds based on feeding strategies taken from Jenkins and Burkhead’s (1993) seminal work on Virginia freshwater fishes, plus an anadromous life cycle group (Table 1). Bulk isotope tissue analysis, elemental analyzer, and isotope ratio mass spectrometry Samples of dorsal muscle tissue were dried at 60°C for three days and homogenized in preparation for analy- sis. The tissues were then lipid extracted by refluxing them in dichloromethane for 35 minutes (Knoff et al., 2002), except for those samples selected for compound specific analysis, which were soxlet extracted (see below; gas chromatography-mass spectrometry (GC- MS) and compound specific stable isotope analysis (CSIA)). One milligram (mg) of dried, lipid-extracted muscle was used for <513C and <515N analysis. Six mg was used for d34S analysis. A Carlo Erba elemental analyzer (EA) (Fisons/VG/Micromass, Manchester, UK) coupled to a Micromass Optima isotope ratio mass spectrom- eter (IRMS) (Fisons/VG/Micromass, Manchester, UK) was used to obtain 613C, 615N and d34S values. The <513C and d15N were obtained concurrently, and <534S was determined during separate analytical runs. MacAvory et al.: Anadromous fish as marine nutrient vectors 167 Figure 1 The boxed area indicates the section of the Rappahannock River, Virginia, between the towns of Fredericksburg and Tappahannock, where all fish were captured to determine the role of anadromous fish as marine nutrient vectors to the freshwater environment. Boat electrofishing was conducted between February and May 1997 and 1999. Sampling was conducted so that fish were captured before, during and after the spring spawning run of the anadromous Alosa spp. The isotope compositions are reported in relation to standard material and follow the same procedure for all stable isotopic measurements, as follows: &E = [(JC£/>'£,)sample/(':£'/>'J5)standard] - 1) x 1000, (1) where E = the element analyzed (C, N, or S); x = the molecular weight of the heavier isotope; and y = lighter isotope te=13, 15, 34, and y=12, 14, 32 for C, N, and S, respectively). The standard materials to which the samples are com- pared are Pee Dee Belemnite for carbon, air N2 for nitrogen, and Canyon Diablo Triolite for sulfur. Repro- ducibility of all measurements was typically 0.2 %c or better. Between every 12 samples, a laboratory standard was analyzed. In a typical run of 60 samples (+5 stan- dards, 65 measurements total) the standard deviations for d15N and 613C were <0.2%e. For <534S, standard devia- tions were <0.3 %o. Gas chromatograph-mass spectrometer (GC-MS) Once dried, muscle samples selected for compound specific isotope analysis (CSIA) were lipid extracted (Soxhlet method from Ballentine et ah, 1996) and the fatty acids had a methyl group added to the carboxyl end (derivitized) so they could be characterized by gas chromatography (GC). This was done by heating with BF3CH3OH for eight minutes (Ballentine et al., 1996). The fatty acid methyl esters (FAME) were analyzed by GC-MS using a Hewlett Packard 5890 Series II gas chromatograph (Palo Alto, CA) interfaced to a Hewlett Packard 5971A mass sensitive detector (Palo Alto, CA), with helium gas as the carrier. A 60-meter J&W DB-5 168 Fishery Bulletin 107(2) column (J&W Scientific, Folsom, CA) was used for FAME separation. The GC oven temperature program used was as follows: 100°C for 2 minutes, ramp at 3°C/min. to 210°C, hold for 20 min, ramp l°C/min. to 220°C, hold for 10 min. Compound specific stable isotope analysis (CSIA) The FAMEs were analyzed for their stable carbon iso- tope compositions using a Hewlett Packard 5890 Series II gas chromatograph interfaced through a combustion furnace with a VG Isoprime IRMS (Fisons/VG/Micro- mass, Manchester, UK). The GC was equipped with the same column that was used for the GC-MS analysis and helium was the carrier gas. The GC oven tempera- ture program was identical to that used for the GC-MS FAME identification. Time elution was used to identify peaks. The C02 combustion products of the fatty acids eluting from the column were introduced into the mass spectrometer after passing through a water trap. All FAME 613C values were corrected for the addi- tion of the methyl group to the original fatty acid. The derivatization of the fatty acids to their methyl esters results in a predictable and reproducible isotope effect (Ballentine et al., 1996; Uhle et al., 1997). Adding a methyl group to the fatty acid alters its isotope signa- ture. However, if the isotopic ratio of the methanol (in this case <513C=-4 6%c, measured by injecting the metha- nol into the mass spectrometer through the GC) and Statistical analysis Kruskal-Wallis nonparametric procedures were used to test for differences in isotopic values among anadromous fish and the dif- ferent guilds (predators, carnivores, general- ists, and planktivores, (a=0.05)). The Dunn procedure was used to examine differences between groups (Rosner, 1990). Statview SE + Graphics (Abacus Concepts, Inc., Cary, NC), JMP In (SAS, Cary, NC) and Microsoft Excel version 5.0 (Microsoft, Inc., Redmond, WA) were used for statistical tests. The Dunn procedure reduces the risk of type-1 error inherent in multiple comparison tech- niques. It does so by increasing the Z-score needed to reject the null hypothesis as the number of individual groups being compared increases. In the present study, a Z-score of ±3.02 (0.9975 confidence) was needed for a difference to be significant. Results The first objective of this study was to estab- lish that the spawning anadromous fish retained the marine isotope signal more than 40 km upstream from saline waters. This was the case for all three isotopes examined. Table t Fish species examined by guild (including an anadromous group) from the Rappahannock River to assess the role of marine fish as nutrient vectors. Guild assignments are based on diet as reported in Jenkins and Burkhead (1993). Guild Species name Common name Predator Ictalurus furcatus blue catfish Lepisosteus osseus longnose gar Carnivore Micropterus salmoides largemouth bass Lepomis gibbosus pumpkinseed Hybognathus regius eastern silvery minnow Notemigonus crysoleucas golden shiner Lepomis macrochirus bluegill Perea flavescens yellow perch Generalist Anguilla rostrata American eel Ameiurus catus white catfish Ameiurus nebulosus brown bullhead Ictalurus punctatus channel catfish Planktivore Menidia beryllina inland silverside Dorosoma cepedianum gizzard shad Erimyzon oblongus creek chubsucker Anadromous Alosa aestivalis blueback herring Alosa pseudoharengus alewife Alosa sapidissima American shad Morone saxatilis striped bass Morone americana white perch the fatty acid methyl ester are known, then the isotopic signature of the original fatty acid can be determined using a mass balance Equation 2. ^13C FAME ~ /FA— <513C FA + /"Methanol Methanol *2) where #3CFAME, 513CFA, and 513C Methanol = the carbon isotope signa- tures of the FAME, the underivatized fatty acid, and the methanol, respec- tively; and fFA and f Methanol = the fraction® of carbon in the FAME due to the underiva- tized fatty acid and metha- nol, respectively (Ballentine et al., 1996; Uhle et al., 1997). Each sample was injected four to eight times (depending on the reproducibility of the analysis). Only d13C values that were within 1.5%e of each other were considered to reflect the 613C of the FAME (MacAvoy et al., 2002). Therefore, the fiL3C reported for each FAME identified is represented by an average value and a standard devia- tion. Every sixth sample injected was an internal, labo- ratory standard (naphthalene-d, 613C-25.7 %c) to insure consistent performance of the GC, oxidation furnace, and mass spectrometer. MacAvory et al.: Anadromous fish as marine nutrient vectors 169 Table 2 Isotope values for all fish used in this study seperated by Family. “A” indicates anadromous, * indicates euryhaline range. Guild assingments are based on diet as reported in Jenkins and Burkhead (1993). “C” indicates a group with some isotope data derived from MacAvoy et al. (2000). White perch ( Morone americana) shows elevated 13C content is probably not marine protein given the low 634S ratio; M. americana is a secondary carnivore and the high 613C reflect this. Standard deviation is given after the ± and N is in parentheses. Family and Species Common name Guild: food types 613C 615N 634S Anguillidae Anguilla rostrata American eel generalist: insects, snails, fish, clams -24.7±0.7 (3) 11.2±0.8 (3) 0.9+2. 4 (3) Atherinidae Menidia beryllina inland sliverside planktivore -23.8±0.9 (3) 15.5 + 0.2 (3) 10.0±0.9 (3) Catostomidae Erimyzon oblongusc creek chubsucker planktivore: planktonic crustaceans -28.1 (1) 10.9 (1) 5.1 (1) Centrarchidae Micropterus salmoides smallmouth bass carnivore -23.0±1.9 (5) 14.5±1.3 (5) 7.6±3.2 (5) Lepomis gibbosus pumpkinseed carnivore: insects, worms -25.4±1.1 (8) 13.1±1.3 (8) 6.5+2. 3 (9) Lepomis macrochirus bluegill carnivore: insects, worms -23.7±2.2 (5) 14.7+1.8(5) 4.7±2.0 (5) Clupeidea Alosa pseudoharengusA ■ c alewife spawning anadromous: copepods, diatoms, ostracods, shrimp, fish — 17.4±1.1 (7) 12.8±0.8 (7) 17.9±0.8 (6) Alosa aestivalis A> c Alosa sapidissimaA< c blueback herring spawning juvenile American anadromous: copepods, cladocerans anadromous: copepods. -19.0±0.6 (7) 13.2±0.3 (7) 17.5±0.4 (7) shadspawning small invertebrates -20.2±0.6 (4) 12.6±0.4 (4) 8.0±2.2 (4) Dorosoma cepedianum gizzard shad planktivore: filter feeder -20.2±2.1 (7) 14.0±0.9 (7) 7.8±2.5 (7) Cyprinidae Hybognathus regius eastern silvery minnowcarnivore: diatoms, algae, ooze detritus -23.0±2.1 (6) 12.4±3.4 (6) 6.5±2.5 (6) Notemigonus crysoleucas golden shiner carnivore: microcrustaceans insects -24.8±1.1 (5) 13.1±1.6 (5) 2.5±1.7 (5) Ictaluridae lctalurus furcatusc blue catfish carnivore/piscivore -21.6±1.9 (43) 15.4±2.0 (43) 9.2±3.0 (43) Ictalurus punctatus channel catfish opportunistic generalist -20.5±2.0 (3) 13.4±1.2 (3) 8.5±3.2 (3) Ameirus nebulosus brown bullhead generalist/omnivorous -24.0±0.8 (3) 13.2±0.5 (5) 5.3±1.6 (5) Ameirus catus white catfish generalist/omnivorous -21.2±2.7 (10) 15.8±2.3 (10) 8.7±4.7 (10) Lepisosteidae Lepisosteus osseus longnose gar predator, piscivore -23.1 16.8 8.34 Moronidae Morone saxatilisA Morone americanaA* striped bass white perch generalist, piscivorous carnivorous: worms, shrimp, fish -25.0±2.3 (2) -20.7±1.2 (5) 13.3±2.4 (2) 16.7 ±1.4 (5) 3.4±4.3 (2) 7.5±3.9 (5) Percidae Perea flavescensc yellow perch carnivore: insects small fish -25.1±2.1 (6) 14.3±2.2 (6) 6.9±1.6 (6) The second objective was to test whether the different guilds of fish showed the incorporation of the marine isotope signal brought to the tidal freshwater by the anadromous fishes. This was observed, but largely lim- ited to the predator guild. Of the groups examined, the anadromous fish were the most 13C-enriched, with mean values of approxi- mately -19%e, followed by predators and planktivores (means -21.8%c and -22.0%e, respectively), which were not significantly different from each other. This sug- gests that, of the remaining two guilds, carnivores were significantly 13C -depleted relative to generalists (mean -24.1%c and -23.5%c, respectively; Table 2). There was approximately a 10%c range in 613C among the exclu- sively freshwater guilds (Table 2, Fig. 2). Anadromous fish have elevated d15N values relative to freshwater fish with similar feeding strategies. However, the trophic enrichment and diet-tissue discrimination associated with <515N signatures make using nitrogen a less effective tracer for source than carbon or sulfur. In this study there was less variability within the guilds 515N signatures, relative to r>13C, although the range (%c) 170 Fishery Bulletin 107(2) H generalist <0 planktivore ▲ carnivore G predator □ anadromous * □ O o °<§P 0°0cfD 0° ° * * © A° A« % ffO «D AD *° *f." a ■ £* m m * ■ * o □ □ cP A. Do dd o D C3 production N 813C(%°) ~ I -20 marine primary production Autochthonous " production Figure 2 815N vs. <513C values for the four guilds and anadromous Alosa spe- cies, showing that most resident freshwater fishes are approximately two trophic levels above primary producers (C3 or autochthonous production), in contrast to the Alosa spp., whose <515N reveals that they are one trophic level above marine primary production. Boxes indicate the isotope signature of C3 terrestrial plant primary pro- duction, freshwater autochthonous production, and marine primary production. Alosa spp. are 13C-enriched relative to most freshwater residents, reflecting marine primary production. 20 15 _ 10 o c o -t 0 -5 -30 -25 -20 -15 813C(%o) Figure 3 <534S vs. 813C values for the four guilds and anadromous Alosa spe- cies, with boxes to indicate the isotope signature of C3 terrestrial plant primary production, freshwater autochthonous production, and marine primary production. Alosa spp. are highly 34S-enriched relative to most freshwater residents, reflecting marine sulfate (which becomes incorporated into primary producers and Alosa spp. while they grow in the Atlantic Ocean). Predators are the only guild showing elevated <534S, indicating the incorporation of marine protein derived from Alosa spp. ■ generalist * planktivore A carnivore G predator □ anadromous Marine protein O O * A O □ „ aD a DIb (5 n D □ o A O CP A° A A □ A O □ . o o & n ■ /-V OA 0®, C3 production Autochthonous production of <515N values among all fishes was similar to that observed for 613C (10%c). The anad- romous fish had the lowest <515N values and generally grouped between 12%c and 13%o; however, their values were not lower than generalists or carnivores. The predators were the most 15N-enriched of any group (Table 2). There were no significant differences among the d15N values for carnivores, generalists, and planktivores (Table 2). Sulfur isotopes were hypothesized to be the most useful for tracing marine protein into freshwater, owing to extreme differ- ences between the <534S of marine plankton and various sulfur sources in freshwater. Predator fishes and anadromous Alosa spp. showed elevated 34S signals relative to other resident freshwater fishes, indicating that the predators incorporated Alosa spp. sulfur (protein). The range of <534S values among all the fish captured was from approximately 0 %o to 20%c, a considerably larger range than observed for the other two isotopes (Table 2, Fig. 3). Significant differences were ob- served in 634S among several of the separate groups. Anadromous species were highly 34S- enriched relative to all resident freshwater fish (Table 2, Fig. 2), although the striped bass (40 cm total length (TL)) had values between 0.3%c and 6.4%c, the lowest of the anadromous <534S values. Predators were the most 34S-enriched of the resident fish, fol- lowed by planktivores (a trend also observed for d13C). Carnivores and generalists were the most 34S-depleted of the guilds and were not significantly different from each other (Table 2). Sulfur was the only stable isotope that completely separated the anadromous Alosa spp. from the full time freshwater resi- dents. All of the Alosa spp. individual values were 34S-enriched and outside the ranges observed in the other groups (Table 2). Fatty acid analysis Fatty acid (FA) isotope values show that some predators derive fats from anadromous fish and that there is a large variation among FA isotope values. FA <513C values were deter- mined for one alewife (anadromous), one giz- zard shad ( Dorosoma cepedianum, a native freshwater planktivore), and two blue catfish ( Ictalurus furcatus, an introduced piscivorous predator). For the blue catfish bulk <513C and 634S values from muscle tissue showed that one individual (A in Table 3) was significantly 13C and 34S-depleted relative to the other. This was also the case for the respective 613C values of their individual FAs. The anadro- mous alewife and the more 13C-enriched blue MacAvory et al.: Anadromous fish as marine nutrient vectors 171 Table 3 Fatty acid (FA) 613C values for Rappahannock River fish. Means ± 1 Standard Deviation. (n=3). Values are corrected for CH40H derevitization. FAs show that carbon from anadromous fish has been incorporated by Ictalarus furcatus but not by other resident fishes. Bulk isotope values show trends similar to the FAs and are as follows: alewife A. pseudoharengus, 613C -19.3%e, 615N 11.9%c, 634S 17.1%e; blue catfish Ictalarus furcatus (A) dL3C -26.0%e, d15N 13.3%e, 634S 6.1%o; 7. furcatus (B) 613C -19.3%e, 615N 16.6%o, d34S 10.8%o; gizzard shad Dorosoma cepedianum 613C -21.5%e, d15N 14.5%o, 634S 10.2%c. Fatty acid Alosa pseudoharengus alewife ( %c ) Ictalurus furcatus blue catfish (%0.05) for any of the other habitat types. 15 20 25 30 35 40 45 Figure 3 Average taxonomic distinctness ( A+ ) of fish assemblages relative to the mean 4+ (dashed line) and the 95% confidence intervals (solid lines) by habitat (A) and management area (open = any area outside the OHAPC open to fishing, OHAPC = areas where all bottom gear except hook and line are restricted, i.e., excluding the OECA, and OECA = inside the MPA where all bottom gear, including hook and line fishing, are restricted) (B) from remotely operated vehicle (ROV) transects conducted on the Oculina Bank during April-May 2003 and October 2005. Habitat assessment : Analysis of digital stills revealed the highest percentage of live coral habitat was found in the OECA making up only 1.9% of the total habitat observed (Fig. 5). A total of 1307 digi- tal still images were taken in 2003 and 2005 and used for analysis. There was significantly more live O. varicosa located within the OECA compared to the OHAPC and open (One-way ANOVA, P= 0.025). The percentage of rock out- crops was significantly higher in the OHAPC compared to the open and OECA as well as in the open compared to the OECA (One-way ANOVA, P<0.001). Significantly more rubble was found in the OECA and open compared to the OHAPC (One-way ANOVA, PcO.OOl). The percentage of pavement was significantly higher in the OECA and OHAPC compared to the open area (One-way ANOVA, P=0.003) and, finally, there was significantly more standing dead O. varicosa in the OECA than the open (One-way ANOVA, P=0.032). Location of video transects and digital still images containing live O. varicosa are shown in Figure 6. Discussion This is the first study to address the functional- ity of coral habitat and to compare fish assem- blages among areas with different management levels on the Oculina Bank. Prior to this study, the last survey conducted on the Oculina Bank was in 2001 (Koenig et ah, 2005), however, several differences exist between the two and new findings have emerged from the current survey. Koenig et al. (2005) targeted high relief sites within the OECA, used side-scan sonar to locate sites, and compared fish densities among three general habitat types (no coral, sparse live and dead O. varicosa, and dense live and dead O. varicosa ). The current study had updated multibeam maps to target sites, Harter et at: Assessment of fish populations and habitat on Oculina Bank 203 compared areas not only within the OECA but also included the OHAPC and open areas, and examined an expanded range of habitats. While it is well known that deep coral habitat supports a high diversity and densities of fish species (Costello et al., 2005; Koenig et al., 2005; Parrish, 2006; Stone, 2006; Ross and Quattrini, 2007), it is unclear whether fish are attracted to live coral or just structure made by corals. Our study addressed this question by comparing fish assemblages, densities, and diversity among several structure-forming habitat types includ- ing coral. We found no significant difference in the composition of fish assemblages or diversity among all hardbottom habitat types. Grouper densities were significantly higher on the most structurally complex habitats (live O. varicosa, standing dead O. varicosa, and rock outcrops) compared to the less complex ones (pavement and rubble). Therefore, higher grouper densities were not exclusive to coral habitats. Accord- ing to Auster (2005), one of the ways to define functionally equivalent habitats is those that support a similar density of fishes, therefore, we conclude that O. varicosa was functionally equivalent to the other hardbottom habitats on the Oculina Bank. Similar results were found in the Gulf of Maine (Auster, 2005). No difference in fish communities was found between habitats dominated by dense corals and those dominated by dense epifauna with or without corals. In addition, Tissot et al. (2006) concluded that fishes in south- ern California were associated with sponges and corals, but no functional relationship was pres- ent. In Hawaii, fish densities were higher in areas with deep-water corals, but when bottom relief and depth were accounted for, these densities were not higher than those for surrounding areas without corals (Parrish, 2006). Ross and Quattrini (2007) concluded that deep slope reefs function much like shallow corals reefs, hosting a unique, probably obligate, ichthyofauna, however other hardbottom habitats were not examined. Even though our study demonstrated that O. varicosa serves a similar role for fishes as other hardbottom habitats, corals are still important and are major contributors to deep-sea habitat complexity and structure (Roberts et al., 2006). Significant numbers of gag and scamp aggregate on and use O. varicosa for spawning habitat and juvenile speckled hind use the coral for shelter suggesting a nursery value of the coral (Gilm- ore and Jones, 1992; Koenig et ah, 2000; Koenig et al., 2005). Intact coral is not only valuable for fish, but invertebrates as well. As long as the coral is standing (live or dead), living space within the colony branches supports dense and diverse communities of associated invertebrates (Reed et al., 2002a, 2002b; Reed et al., 2007). However, once reduced to unconsolidated coral rubble, little living 50-, 45 40- 35 30- 25- 20- 15- 10- l ill pavement rubble rock outcrops standing dead live O. varicosa O. varicosa Figure 4 Average grouper densities (no. /hectare) (±SE) for each man- agement area by habitat type observed from remotely operated vehicle (ROV) transects conducted on the Oculina Bank during April/May 2003 and October 2005. Average grouper density for pavement in the open area was 0.0 fish/hectare, however, there were no live or standing dead Oculina varicosa transects for the open area. 70- 60- o) 50 - 40- a> cn co 5 > < 30- 20- 10- [ii rf pE ig .Jfl L * * pavement rubble rock outcrops standing dead O. varicosa live O. varicosa Figure 5 Average percent cover (±S.E.) of habitat types in each of the three management areas (open = any area outside the OHAPC open to fishing, OHAPC = areas where all bottom gear except hook and line are restricted, i.e., excluding the OECA, and OECA = inside the MPA where all bottom gear, including hook and line fishing, are restricted) from analysis of digital stills taken during remotely operated vehicle (ROV) transects on the Oculina Bank during April-May 2003 and October 2005. space is left except for infauna (George et al., 2007). A hypothetical trophic model of the O. varicosa ecosystem indicates significant loss of habitat, in particular intact 204 Fishery Bulletin 107(2) live and dead standing coral, could bring dramatic shifts in the ecosystem (George et al., 2007). Conserva- tion efforts, however, should focus on the intrinsic value of corals such as their slow growth, high sensitivity to disturbance, and questionable potential for recovery (Auster, 2005). A restoration project utilizing artificial reef structures is currently ongoing within the OECA. Between 1996 and 2001, a total of 125 large and 900 small restoration modules were deployed in a series of experiments to test their efficacy in the recovery of degraded coral and depleted fish populations (Koenig et al, 2005). The theory is that this will help O. vari- cosa restoration by providing stable settlement habitat, which may, in turn, provide suitable habitat for fish populations to recover. Early evidence (ROV dives from this study) found new coral recruits growing on the structures and groupers associated with them as well (Reed et al., 2005). While the scale of the artificial reefs is likely too small for fisheries replenishment, this experiment will provide insight to whether this tool is effective for coral restoration. 80°15'0"W 80°0'0"W 79°45'0"V^ Locations of live Oculina varicosa (ivory tree coral) from video and digital stills collected during remotely operated vehicle (ROV) transects during April-May 2003 and Octo- ber 2005. Being the first study to compare fish assemblages among areas with different management levels on the Oculina Bank, the results are important to the South Atlantic Fishery Management Council as they evalu- ate the effectiveness of the OECA; this study and fu- ture surveys will help determine the fate of the closed area when it is reconsidered by the Coral and Habitat Advisory Panels in 2014. While MDS and ANOSIM analyses revealed no significant differences in the com- position of fish assemblages among management areas, other positive effects of the closure were observed. Fish diversity was higher inside the OHAPC and OECA compared to the open area. Grouper densities were significantly higher in the OECA, particularly on rock outcrops, than in the OHAPC or open areas. Also, more coral was found in the OECA suggesting the restriction of fishing activity may have aided in conserving what little O. varicosa had not been destroyed by trawling. Habitat quantification analyses demonstrated there was significantly more live and standing dead O. vari- cosa in the OECA compared to the OHAPC and open. An important observation from the ROV transects was the presence of black sea bass ( Centropristis striata) in 2005. Prior to that time, black sea bass had not been observed on the O. varicosa reefs since the 1980s when they dominated the area (Koe- nig et al., 2000). While black sea bass in the 1980s were large, mature individuals, most individuals in 2005 were small ju- veniles, ranging in length from 10 to 20 cm, suggesting initial stages of recovery for this species. Another significant dis- covery was the sighting of the first juvenile speckled hinds since the 1980s. All of these findings combined present initial evidence demonstrating effectiveness of the MPA for restoring reef fish and their habitat. Sustained enforcement remains an on- going problem for MPAs (Riedmiller and Carter, 2001; Rogers and Beets, 2001). Even relatively moderate levels of poach- ing can quickly deplete gains achieved by closure (Roberts and Polunin, 1991; Russ and Alcala, 1996). As of 2003, all trawl- ing vessels working in the Oculina Bank area are required to have vessel monitoring systems, but this doesn’t solve the problem of poaching by hook and line fishing. Be- tween 2003 and 2007, illegal trawlers and fishers were observed within the MPA dur- ing our cruises, and several vessels have been cited and fined by the United States Coast Guard. ROV observations from this study indicate recent trawl nets, bottom long lines, and fishing lines inside the MPA long after these gears were banned from the area. Continued trawling and bottom fishing in the OHAPC likely will thwart management objectives. Harter et al. : Assessment of fish populations and habitat on Oculina Bank 205 In summary, unlike shallow-water ecosystems, un- derstanding of the ecological and functional role of deep-water corals has only recently emerged. The cur- rent study is in agreement with most other recent lit- erature, demonstrating that corals are functionally equivalent to other deep-sea structural habitats. Deep- sea corals, however, are clearly an important provider of structural habitat for fishes and are sensitive to fishing gear impacts and vulnerable to destruction due to their fragility and slow growth rates. There- fore, protection remains crucial. While an ecosystem approach to management has become widely accepted and MPAs have become a primary tool to manage deep- sea coral ecosystems, little evidence has been provided demonstrating MPA effectiveness. This study, however, revealed several positive effects of the closure including higher biodiversity, grouper densities, and percentage of intact coral suggesting initial effectiveness of the Oculina MPA. Acknowledgments We thank the United Space Alliance (USA), National Aeronautics and Space Administration (NASA), and the crews of the MV Liberty Star and MV Freedom Star for providing vessel support. L. Horn and G. Taylor of University of North Carolina at Wilmington/National Underwater Research Center (UNCW/NURC) provided ROV support. Funding was provided by National Ocean- ographic and Atmospheric Administration ( NOAA) Office of Exploration, National Marine Fisheries Service South- east Fisheries Science Center (NMFS SEFSC), NOAA Coral Reef Conservation Program (CRCP), and NOAA Undersea Research Center at UNCW. M. Miller of NMFS SEFSC acted as NOAA CRCP representative. A. Maness of UNCW/NURC provided the image analysis program. We also thank C. Koenig and S. Brooke whose comments helped to improve the manuscript. Literature cited Allison, G. W., J. Lubchenco, and M. H. Carr. 1998. Marine reserves are necessary but not sufficient for marine conservation. Ecol. Appl. 8:S79-S92. Auster, P. J. 2005. Are deep-water corals important habitats for fishes? In Cold-water corals and ecosystems (A. Freiwald and J. M. Roberts, eds.), p. 747-760. Springer, Berlin, Heidelberg. Avent, R. M., M. E. King, and R. H. Gore. 1977. Topographic and faunal studies of shelf-edge promi- nences off the central eastern Florida coast. Hydrobiol 62(2):185-208. Bohnsack, J. A. 1998. Application of marine reserves to reef fisheries management. Aust. J. Ecol. 23:298-304. Clarke, K. R. 1993. Non-parametric multivariate analyses of changes in community structure. Aust. J. Ecol. 18:117-143. Clarke, K. R. and R. M. Warwick. 1998. A taxonomic distinctness index and its statistical properties. J. Appl. Ecol. 35:523-531. 2001. Change in marine communities: an approach to statistical analysis and interpretation, 2nd ed. PRIMER- E, Plymouth, U.K. Costello, M. J. M. McCrea, A. Freiwald, T. Lundalv, L. Jonsson, B. J. Bett, T. C. E. van Weering, H. de Haas, J. M. Roberts, and D. Allen. 2005. Role of cold-water Lophelia pertusa coral reefs as fish habitat in the NE Atlantic. In Cold-water corals and ecosystems (A. Friewald and J. M. Roberts, eds.), p. 771-805. Springer-Verlag, Berlin, Heidelberg. Fossa, J., P. Mortensen, and D. Furevik. 2002. The deep-water coral Lophelia pertusa in Norwe- gian waters: distribution and fishery impacts. Hydro- biol. 471:1-12. George, R. Y., T. A. Okey, J. K. Reed, and R. P. Stone. 2007. Ecosystem-based fisheries management of seamount and deep-sea coral reefs in U. S. waters: conceptual models for proactive decisions. In Conservation and adaptive management of seamounts and deep-sea coral ecosystems (R. Y. George and S. D. Cairns, eds.), p. 9-30. Rosentiel School of Marine Science and Atmo- spheric Science, Univ. Miami, Miami, FL. Gilmore, R. G. and R. S. Jones. 1992. Color variation and associated behavior in the epinepheline groupers, Mycteroperca microlepis (Goode and Bean) and M. phenax Jordan and Swain. Bull. Mar. Sci. 51( 1):83-103. Guenette, S., T. Lauck, and C. Clark. 1998. Marine reserves; from Beverton and Holt to the present. Rev. Fish Biol. Fish. 8:251-272. Husebo, A., L. Nottestad, J. H. Fossa, D. M. Furevik, and S. B. Jorgensen. 2002. Distribution and abundance of fish in deep coral habitats. Hydrobiol. 471:91-99. Jonsson, L. G., P. G. Nilsson, F. Floruta, and T. Lundaelv. 2004. Distributional patterns of macro- and megafauna associated with a reef of the cold-water coral Lophelia pertusa on the Swedish west coast. Mar. Ecol. Prog. Ser. 284:163-171. Koenig, C. C., F. C. Coleman, C. B. Grimes, G. R. Fitzhugh, K. M. Scanlon, C. T. Gledhill, and M. Grace. 2000. Protection of fish spawning habitat for the conser- vation of warm-temperate reef-fish fisheries of shelf-edge reef of Florida. Bull. Mar. Sci. 66(3):593-616. Koenig, C.C., A. N. Shepard, J. K. Reed, S. D. Brooke, J. Brusher, and K. M. Scanlon. 2005. Habitat and fish populations in the deep-sea Ocu- lina coral ecosystem of the western Atlantic. Am. Fish. Soc. Sym. 41:795-805. Kohler, K. E. and S. M. Gill. 2006. Coral point count with Excel extensions (CPCe): a visual Basic program for the determination of coral and substrate coverage using random point count methodology. Computers and Geosciences 32(9):1259- 1269. Koslow, J. A., G. W. Boehlert, J. D. M. Gordon, R. L. Haedrich, P. Lorance, and N. Parin. 2000. Continental slope and deep sea fisheries: impli- cations for a fragile ecosystem. ICES J. Mar. Sci. 57:548-557. Parrish, F. A. 2006. Precocious corals and subphotic fish assemb- lages. Atoll Res. Bull. 543:425-438. 206 Fishery Bulletin 107(2) Pyle, R. L. 2000. Assessing undiscovered fish biodiversity on deep coral reefs using advanced self-contained diving technology. Mar. Tech. Soc. J. 34:82-91. Reed, J. K. 1980. Distribution and structure of deep-water Oculina varicosa coral reefs off central eastern Florida. Bull. Mar. Sci. 30(3):667-677. 2002a. Deep-water Oculina coral reefs of Florida: biology, impacts, and management. Hydrobiol. 471:43—55. 2002b. Comparison of deep-water coral reefs and litho- herms off southeastern USA. Hydrobiol. 471:57-69. Reed, J. K., C. C. Koenig, and A. N. Shepard. 2007. Impacts of bottom trawling on a deep-water Oculina coral ecosystem off Florida. Bull. Mar. Sci. 81:481-496. Reed, J. K., A. N. Shepard, C. C. Koenig, K. M. Scanlon, and R. G. Gilmore Jr. 2005. Mapping, habitat characterization, and fish surveys of the deep-water Oculina coral reef marine protected area: a review of historical and current research. In Cold-water corals and ecosystems: proceedings of the second international symposium on deep sea corals (A. Freiwald, J. Roberts, eds.), p. 443-465. Springer-Verlag, Berlin Heidelberg. Riedmiller, S. and E. Carter. 2001. The political challenge of private sector management of marine protected areas: The Chumbe Island Case, Tanzania. ACP-EU Fish. Res. Rep. no. 10, p. 141-153. Roberts, C. M. 2002. Deep impact: the rising toll of fishing in the deep sea. Trends Ecol. Evol. 17:242-245. Roberts, C. M. and N. V. C. Polunin. 1991. Are marine reserves effective in management of reef fisheries? Rev. Fish Biol. Fish. 1:65-91. Roberts, J. M., A. J. Wheeler, and A. Friewald. 2006. Reefs of the deep: the biology and geology of cold- water coral ecosystems. Science 312:543-547. Rogers, C. S. and J. Beets 2001. Degradation of marine ecosystems and decline of fishery resources in marine protected areas in the US Virgin Islands. Environ. Conserv. 28(4):312- 322. Ross, S. W. and A. M. Quattrini. 2007. The fish fauna associated with deep coral banks off the southeastern United States. Deep-Sea Research I. 54:975-1007. Russ, G. R. and A. C. Alcala. 1996. Marine reserves: rate and patterns of recovery and decline of large predatory fish. Ecol. Appl. 6:947-961. Stone, R. P. 2006. Coral habitat in the Aleutian Islands of Alaska: depth distribution, fine-scale species associations, and fisheries interactions. Coral Reefs 25:229-238. Tissot, B. N., M. M. Yoklavich, M. S. Love, K. York, and M. Amend. 2006. Benthic invertebrates that form habitat on deep banks off southern California, with special reference to deep sea coral. Fish. Bull. 104:167-181. Watling, L. and E. A. Norse. 1998. Disturbance of the seabed by mobile fishing gear: a comparison to forest clearcutting. Conserv. Biol. 12(6):1180-1197. 207 Changes in body composition and fatty acid profile during embryogenesis of quillback rockfish ( Sebastes maSiger) Cara J. RodgveSSer Email address for contact author: Fletcher.Sewall@noaa.gov National Marine Fisheries Service, NOAA Alaska Fisheries Science Center Auke Bay Laboratories Ted Stevens Marine Research Institute 17109 Pt. Lena Loop Road Juneau, Alaska 99801 Abstract — We investigated develop- mental changes in the body composi- tions and fatty acid (FA) profiles of embryos and preparturition larvae of the quillback rockfish ( Sebastes maliger). Comparisons of proximate composition data from early-stage embryos with data from hatched preparturition larvae taken from wild-caught gravid females indi- cated that embryos gain over one- third their weight in moisture while consuming 20% of their dry tissue mass for energy as they develop into larvae. Lipid contributed 60% of the energy consumed and was depleted more rapidly than protein, indicating a protein-sparing effect. Oil globule volume was strongly correlated with lipid levels, affirming its utility as an indicator of energetic status. FA profiles of early embryos differed significantly from those of hatched larvae. Differences in the relative abundances of FAs between early embryos and hatched larvae indicated different FA depletion rates during embryonic development. We conclude that some metabolically important FAs may prove useful in assessing the condition of embryos and prepar- turition larvae, particularly 20:4n-6, which cannot be synthesized by many marine fish and which is conserved during embryogenesis. Variability in body composition and energy use among rockfish species should be considered when interpreting any measures of condition. Manuscript submitted April 14, 2008. Manuscript accepted November 19, 2008. Fish. Bull. 107:207-220 (2009). The views and opinions expressed or implied in this article are those of the author and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Fletcher F. Sewall (contact author) The nutritional condition of fish during their early life histories may play a major role in determining the strength of year classes because larvae must have energy stores sufficient to ensure survival to first feeding. The survival rates of early planktonic rockfish larvae may be influenced by differences in the amounts and use of endogenous protein and lipid sources during embryonic develop- ment (MacFarlane and Norton, 1999). Despite this potential importance, little is known about the biochemis- try of developing rockfish embryos and larvae. Because utilization of lipid and protein may vary by species (e.g., MacFarlane and Norton, 1999) and life history stage (e.g., Norton et al., 2001), it is important to exam- ine these variables by species at the appropriate life stage. Quillback rockfish (Sebastes ma- liger) are a long-lived, slow-growing species of commercial importance, for which biochemical data on early life stages are lacking. Like other rock- fish species of the genus Sebastes, they bear live young, and embryos (as post fertilization, prehatching in- dividuals with the chorion intact) de- velop and hatch as larvae (individuals free of the chorion envelope) inside the maternal female before being ex- truded (Yamada and Kusakari, 1991). Survival during the larval phase can be vital in determining the eventu- al size of a rockfish cohort (Ralston and Howard, 1995). The utilization of lipids is of particular importance, as triacylglycerols (TAGs) and polar lipids (mainly phospholipids) may be the primary energy sources during rockfish embryogenesis (MacFarlane and Norton, 1999). Endogenous TAG is thought to reside mainly in an oil globule, the volume of which was iden- tified as a main correlate of survival of black rockfish larvae (S. melanops ) (Berkeley et al., 2004). In that study, total lipid concentration was not re- lated to oil globule volume (OGV) or later larval survival; however, lipid levels have been correlated with sur- vival for many other fish species (re- viewed in Kamler, 1992). Research with wild-caught shortbelly rock- fish (S. jordani) (pre-flexion larvae through juvenile stages) has indicated that the relationship of TAG to total lipids, and the usefulness of TAG as an indicator of nutritional status, depends upon life stage (Norton et al., 2001). Given this variability, it is unclear what trends may occur in total lipid levels and oil globule TAG reserves in developing quillback rock- fish embryos. If OGV can be shown to be a reliable indicator of lipid levels, using this measurement would rep- resent a substantial savings in time and cost as compared with analytical chemistry techniques. Embryos and larvae of quillback rockfish are likely incapable of syn- thesizing essential fatty acids (EFAs), either entirely or at a rate which will meet their metabolic needs for growth and survival, as is the case for adults of other fish species (e.g., as reviewed 208 Fishery Bulletin 107(2) Table 1 Sample sizes for determinations of body composition and fatty acid profiles of quillback rockfish ( Sebastes maliger) embryos (early and middle stages) and hatched, preparturition larvae (late stages). The sample unit was one maternal female, from which sub- samples of embryos or larvae were obtained for use in biochemical analyses. Sample sizes varied due to inadequate subsample masses being available for some analytical procedures. Variable Sample size Early stages (1-3) Middle stages (4-9) Late stages (10) Developmental stage 5 6 4 Oil globule volume 3 4 4 Wet tissue mass 3 4 4 Moisture, protein 3 4 4 Ash 2 4 3 Lipid 3 4 4 Fatty acids 4 4 4 in Watanabe, 1982). Fish are capable of selectively ca- tabolizing particular fatty acids (FAs) while retaining others (reviewed in Tocher, 2003). Differences in rates of individual FA use during embryogenesis would be reflected by changes in overall FA profiles as embryos develop into hatched larvae. Assessing net differences in the amounts of individual FAs present may reveal which FAs potentially contribute to variability in larval survival (e.g., due to deficiencies in particular EFAs resulting from inadequate maternal provisioning). Our study was driven by three objectives. First, we sought to describe the amount and sources of energy consumed during quillback rockfish embryogenesis, by measuring changes in lipid and protein levels from early to late stages of development. Second, to assess the usefulness of OGV as an indicator of the energetic status of embryos and preparturition larvae, we inves- tigated how well changes in OGV were correlated with changes in stage and biochemical composition. Last, we reduced lipids to their FA components and compared the overall FA profiles of embryos to preparturition larvae, to determine whether all FAs were used at the same rate as the total lipid during embryonic development, or whether some were used disproportionately fast while others were conserved. Methods Sampling Quillback rockfish were caught 15-28 April 2006 by hook and line in southeastern Alaska on the northwest side of Chichagof Island (58°10'N, 136°21'W). Fish were caught within approximately 1 km of shore at depths of 30 to 75 m. Fifteen gravid females ranging in size from 360 to 480 mm (fork length) were transported live to Auke Bay Laboratory in Juneau, where they were kept in flow-through seawater tanks at 3.5-4°C. During a 2-week holding period, the females did not feed and did not release larvae naturally. Females were then sacrificed and a sample of embryos or larvae was manu- ally expressed from each fish. Sample sizes available for biochemical analysis varied occasionally because each analytical procedure was destructive and required separate subsamples of embryos or larvae, and sample masses were below the minimum needed to ensure accurate analysis in some cases (Table 1). One sample of stage seven embryos was omitted from analysis due to the apparent degradation and possible resorption of embryos by the parent. Changes in lipid and protein levels during development Developmental stages We ranked embryos or larvae from each female in order of development (stages 1-10, from immediately after fertilization through posthatch- ing; Fig. 1) following the descriptions of kurosoi rockfish (S. schlegelii ) by Yamada and Kusakari (1991), and incorporating our own observations for quillback rockfish (Table 2). In quillback rockfish, we found that the retina went through many stages of pigmentation and that body pigment appeared relatively early in development and became more pronounced through time. Yamada and Kusakari (1991) include only one stage for retinal pigmentation and one for peritoneal pigment (stages 25 and 29, respectively), so we further divided the embryo stages based on these characteristics. Developmental stages were then used for tracking changes in body composition during embryonic devel- opment. Because the durations (in days) of stages vary widely (Eldridge et al., 2002), they are not strictly ap- propriate for statistical analyses with linear models. In any model using developmental stage categories, the assumption that the stages represent equal inter- vals can distort the true patterns of change over time. Quantitative statements about rates of change in body composition ideally would be based on time since fer- Sewall and Rodgveller: Changes in body composition and fatty acid profile during embryogenesis of Sebastes maliger 209 Table 2 Developmental staging scheme for quillback rockfish (Sebastes maliger ) embryogenesis. Stages 1 through 9 represent progres- sively developing embryos, whereas stage-10 samples contained many hatched larvae. Equivalent stages from Yamada and Kusa- kari (1991) are included for comparison. Yamada and Kusakari (1991) Stage Description stage 1 Embryonic shield (very small germ disc on one pole of egg) 15 2 Head fold 16 3 Optic vesicles 17 4 Optic cups, increased orbital definition 20 5 Early retinal pigmentation 25 6 Retinal pigment light, spreading throughout eye; body pigment appears as scattered dark dots along ventral side of tail 25-28 7 Very slight eye shimmer appears; body pigment increased slightly, still ventral 25-28 8 Eye shimmer increases, scattered throughout the darkening retina; body pigment increases >2x, still ventral, spots merging to form a line 25-28 9 Retina dark with a lot of shimmer scattered throughout, some black still visible; pigmentation on gut behind yolk sac and dorsally along tail 25-28 10 Dark retina covered with shimmer, body pigment blended into a dark line on ventral side of tail, spots also on dorsal side of tail and on peritoneal wall; hatched/hatching imminent; yolk not depleted 29-31 Figure 1 Quillback rockfish (Sebastes maliger) stage-1 embryo (left) and stage-10 hatched larva (right). tilization; however, we did not possess data on the gestation period for quillback rockfish. The period of gestation seems to vary widely among rockfish spe- cies (e.g., 29 days for S. flavidus [Eldridge et al„ 2002], 48 days for S. schlegelii [Yamada & Kusakari, 1991]), as well as the time spent at each stage of devel- opment, so we did not feel confident in assigning esti- mated time durations to each stage based on studies of other rockfish. However, we were more concerned with general trends during development, and net differences in body composition as embryos become larvae, than the precise rates of change among stages. In addition, other studies have reported developmental changes in body composition using stages assumed to represent equal intervals (e.g., MacFarlane and Norton, 1999); to facilitate comparisons, we also chose to follow this convention. To assess net changes in body composition that oc- curred over the course of embryogenesis (i.e., differences between early embryos versus hatched, preparturition larvae), data on body compositions were averaged from three samples at the earliest available stages (stages 2 and 3) and compared with values averaged from four late-stage samples (stage 10). To describe trends and variability in lipid and protein use across all stages of development, protein and lipid masses were plotted against developmental stage and the strengths of the correlations were calculated. Samples at stages 1 and 9 were excluded from biochemical analyses due to techni- cal constraints, such as insufficient sample masses for some processes. Wet tissue mass The average wet tissue mass of embryos and larvae at each developmental stage was determined for use in calculations of body composition. A subsample of ~100 embryos or larvae removed from a maternal fish was placed on filter paper to drain intraovarian fluid. The subsample was then weighed to the nearest 0.1 mg, and individuals were counted under a dissecting micro- scope. This was repeated three times per female, and data from the three replicates were used to calculate an average wet mass per embryo or larva. These sub- samples were discarded to prevent degraded or oxidized samples from being included in biochemical analyses. 210 Fishery Bulletin 107(2) The remaining embryos or larvae from a female were placed in a 20 ml glass vial capped with nitrogen and stored in a freezer at -80°C to prevent oxidation and tissue degradation prior to further processing. Moisture, protein, and ash content A subsample of approximately 2-4 g (wet mass) of embryos or larvae, representing a composite of thousands of individuals, was used from each sample for analysis of moisture, protein, and ash (inorganic components such as phos- phorous, calcium, and other minerals). To determine percent moisture, samples were placed in crucibles in a Leco Thermogravimetric Analyzer 601 (TGA 601) (Leco Corporation, St. Joseph, MI), heated to 135°C to boil off moisture, and wet and dry sample masses were com- pared. Percent ash was determined gravimetrically by further heating samples to 600°C to combust all organic components and weighing the remaining mass. The dry mass percent protein was calculated as the observed nitrogen content multiplied by a factor of 6.25, based on the assumption that nitrogen accounts for 16% of the protein mass (Craig et al., 1978). Nitrogen content was determined following the Dumas method (Horwitz, 2002), using a Leco FP 528 nitrogen analyzer (Leco Corporation, St. Joseph, MI), with approximately 0.1 g dry sample mass burned at 850°C and the re- leased nitrogen measured by thermal conductivity. A National Institute of Standards and Technology Standard Reference Material (SRM) 1546 (pork and chicken homogenate) was used to calibrate the Leco TGA 601, and Leco calibration sample ethylenediamine- tetraacetic acid (EDTA, 9.57 ± 0.04% nitrogen) was used to calibrate the Leco FP 528. Two quality assurance samples, Chinook salmon ( Oncorhynchus tshawytscha) homogenate and walleye pollock (Theragra chalcogram- ma) homogenate, were subjected to proximate analysis along with the larval samples to verify the accuracy of protein, moisture, and ash measurements. Repli- cate measurements of nitrogen content were taken as a check for precision, with a target error limit of less than 15% coefficient of variation. Carbohydrate content was not analyzed in this study because fish eggs typically have very low levels of car- bohydrates, averaging 2.6% of dry mass (Kamler, 1992). Adult fish also do not typically store carbohydrates in any appreciable quantities (Brett, 1995). Lipid content A subsample of 0.2 to 0.3 g wet mass containing hundreds of embryos or larvae was used from each of the samples for lipid analysis. Samples were processed by a modified Folch’s method as described by Christie (2003). A 2:1 solution of chloroform and methanol, with 0.1 g/L butylated hydroxytoluene (BHT) to minimize oxidation, was used to extract lipids under high temperature (120°C) and pressure (1200 psi) on a Dionex ASE 200 Accelerated Solvent Extractor (Dionex Corporation, Sunnyvale, CA). Extracts were washed with 0.88% KC1 followed by a 1:1 (by volume) metha- nol/deionized water solution, both added at 25% of the extract volume, to remove co-extractables (e.g., glycerol) from the solution containing the extracted lipids. The resulting extract volume was reduced to less than 1 mL by evaporating excess solvent with a Yamato RE 540 rotary evaporator system (Yamato Scientific America, Inc., Santa Clara, CA), then drawn up by electronic pipette with sufficient chloroform to bring the volume to 1000 qL. For gravimetric analysis of total percent lipid, a 500-pL aliquot of the extract was placed in aluminum weighing pans in a fume hood overnight, allowing the solvent to evaporate and leave behind the extracted lipids. The remaining half of the extract was capped with nitrogen and stored at -80°C to minimize oxidation until further processing for FA analysis. In quality assurance tests, the extraction method con- sistently yielded wet tissue lipid concentration values not exceeding 15% error compared with the certified value for Standard Reference Material 1946 (lake trout [ Salvelinus namaycush]). Three quality control samples were also processed concurrently with the larval sam- ples. A method blank containing no sample was used to verify that any contaminants or residues that could bias the observations of lipid mass were less than 0.01% of the average sample mass. As a check for accuracy, extraction of Pacific herring ( Clupea pallasii ) reference tissue yielded lipid concentrations that varied by less than 8% from the average value established in prior analyses. As a check for precision, one larval sample was split into two portions that yielded percent lipid values with less than 1% coefficient of variation. Energy estimates Total energy content, energy density, and the relative energetic contributions of protein and lipid were esti- mated from protein and lipid masses. Protein mass was expressed as its energy equivalent by calculating the product of protein mass and an energy density of 20.1 J/mg, and a similar calculation was made for lipids using an energy density of 36.4 J/mg — figures which are con- versions of the average energy density values reported by Brett (1995). For samples having both protein and lipid analyses completed, these were combined to estimate the total energy content per individual embryo or larva, and expressed in relation to sample wet and dry masses to obtain energy density values. Oil globule volume Subsamples of 16 to 37 embryos or larvae (mean=24) from each female were placed in Petri dishes and photo- graphed digitally under a dissecting microscope. Using the Clever Ruler 3.0 software (shareware published by zcstar.com), we measured two perpendicular oil globule diameters for each larva from the photos. An average oil globule volume (OGV) for each was then converted to millimeters using a stage micrometer at the same magnification. The change in OGV was determined as the difference in average OGV between early embryonic stage samples and hatched larval samples. To describe trends and variability in OGV across all stages of devel- Sewall and Rodgveller: Changes in body composition and fatty acid profile during embryogenesis of Sebastes maliger 211 opment, OGV was plotted against developmental stage and the strength of the correlation calculated. The rela- tionship between OGV and energetic status of larvae was assessed by treating lipid mass, lipid concentration, and protein mass as response variables and OGV as a predictor variable in simple linear models. Significance tests were performed with a one-way analysis of vari- ance (ANOVA). Fatty acid analysis FA composition of total lipid extracts was determined by gas chromatography and mass spectrometry. To pre- pare lipid extracts for FA analysis, whole lipid extracts underwent acid-catalyzed transesterification to fatty acid methyl esters (FAMEs), following a procedure out- lined by Christie (2003). Two mL of Hilditch reagent (0.5 N sulfuric acid [H2S04] in methanol) was added to an aliquot of lipid extract which contained 0.3 mg of lipid. Before transesterification, 2050 nanograms (ng) of 19:0 FA in 50.0 pL hexane was added to each sample as an internal standard for quantification. The solution was incubated at 55°C for approximately 18 hours, and then washed with 5 mL of 5% aqueous sodium chlo- ride (NaCl). To separate and extract the FAMEs from the aqueous solution, 4 mL of hexane was added, the solution was stirred on a vortex mixer, and the hexane layer transferred by pipette to a second container; this process was repeated with another 4 mL of hexane. Four milliliters of 2% potassium bicarbonate (KHC03) was added to the hexane containing the FAMEs to quench the esterification reaction and neutralize any remaining acid. The hexane-FAME layer was run through a sodium sulfate (Na2S04) drying column to remove any residual co-extractables and water, and the resulting hexane-FAME volume reduced to approximately 1 mL in a Labconco Rapidvap (Labconco Corporation, Kansas City, MO). Prior to GC analysis, 2040 ng of 21:0 FAME in 50.0 pL hexane as an instru- mental internal standard was added to each sample for use in sample recovery calculations. The FAMEs were then eluted with a temperature gradient on a Hewlett Packard 6890 gas chromatograph (Hewlett-Packard Company, Palo Alto, CA) with a 5973 mass selective detector by using a 30-m Omegawax 250 fused silica column ( Sigma-Aldrich, St. Louis, MO). Five-point calibration curves were created from known concentra- tions of a Supelco FAME-37 standard mix (Supelco, Bellefonte, PA). Thirty of the 32 FAMEs investigated yielded calibration curves with a coefficient of deter- mination r2 > 0.990. As a quality assurance measure, selected calibration standards were re-injected and quantified, and the average across all FAME analytes fell within ±1.5% of the known value. Along with the samples, quality control samples from the lipid extraction step were subjected to the trans- esterification procedure. Concentrations of 23 of the 28 FAMEs detected in the Standard Reference Material 1946 were within 25% of the average values obtained from six previous analyses, with none exceeding 35% error. Duplicate larval samples yielded FA concentra- tions with coefficients of variation less than 10% for 26 of the 29 FAs present. Six FAMEs were detected in the method blank (in order of mass: 18:0, 16:0, 22:ln-9, 17:0, 18:ln-9 cis and trans, and 14:0) and the masses of these were subtracted from the masses of those FAMEs found in each of the samples as a correction. Statistical analysis To determine whether FA profiles of early-stage embryos differed from hatched, preparturition larvae, raw data on FA concentrations (ng of FA per g of wet sample mass) were first converted to proportions of total FAs per sample. The relative proportions of individual FAs present in four samples of early-stage embryos were then compared to those found in four late-stage samples by analysis of similarities (ANOSIM), a nonparametric, multivariate statistical test suitable for compositional studies (Clarke and Warwick, 1994). ANOSIM was per- formed on a dissimilarity matrix based on the Aitchison distance (Aitchison, 1992) between all possible pairs of samples. The Aitchison distance ( DAitchison ) between two samples, A and B, is derived from the differences between the log ratios of pairs of FAs present in the two groups: D Aitchison (A.B) KJ A log A Bj) (1) where j takes on values up to the number of analytes investigated — in this case, 32. The Aitchison distance cannot be calculated in cases where the concentrations of a FA are zero in any of the samples being compared. This proved not to be a significant limitation because only one FA, 18:ln-ll, was present in measurable quantities in some samples but not in others. This FA was excluded from ANOSIM analysis, but included in estimates of FA mass losses. Three other FAs (15:ln-5, 17:ln-7, and 18:2n-6 trans ) were also excluded from analysis because they yielded zero values for all samples. We used ANOSIM to com- pare the ranked Aitchison distances among samples within groups and among samples between groups. This yielded the ANOSIM R statistic, which can range in value from -1 to 1, with a zero value indicating identi- cal groups (i.e., all FAs were used at the same rate, resulting in no difference between FA compositions of embryos and hatched larvae), positive values indicating dissimilarity between groups (i.e., FAs were used at different rates, resulting in changes to the FA compo- sitions of embryos as they developed into larvae), and negative values indicating greater dissimilarity within than between groups (i.e., a study design problem). The significance value was determined through permuta- tions where the observed R value is compared to simu- lated R values assuming no difference between groups 212 Fishery Bulletin 107(2) Table 3 Quillback rockfish ( Sebastes maliger) body composition data averaged for early-stage embryos and hatched preparturition larvae (mean ±1 SD). Comparisons of early versus late-stage samples revealed net changes that occurred during embryogenesis. “Early” included stages 2-3 embryos, and sample size ( n ) was 3 maternal females, except for ash (n= 2). “Late” included stage-10 larvae, n = 4, except ash (n= 3). Comparisons only included those samples for which all proximate composition data, except ash, were available. Each sample was a composite of hundreds of embryos or larvae from the same parent. Dry masses did not sum to exactly 100% because lipid was determined by a separate process from protein and ash. Early Late % Change Wet mass per individual ( pg) 649 ±60 884 ±72 36.2 Moisture (%) 78.5 ±1.7 87.3 ±0.8 11.3 Dry mass per individual (pg) 140 ±13 112 ±15 -19.5 Protein mass per individual (pg) 90.3 ±9.9 73.0 ±12.1 -19.2 Lipid mass per individual (pg) 43.5 ±6.8 28.9 ±4.7 -33.6 Ash mass per individual (pg) 8.61 ±0.57 9.71 ±0.90 12.8 Protein (% wet mass) 13.9 ±1.08 8.22 ±0.74 -41.0 Lipid (% wet mass) 6.70 ±0.85 3.25 ±0.35 -51.4 Lipid (% dry mass) 31.0 ±1.9 25.7 ±2.2 -17.3 Total energy content per individual ( J) 3.40 ±0.44 2.52 ±0.40 -25.9 Energy density ( J/mg wet mass) 5.24 ±0.52 2.84 ±0.24 -45.9 Energy density (J/mg dry mass) 24.3 ±0.9 22.4 ±0.8 -7.9 Oil globule volume (nL) 27.8 ±1.2 13.7 ± .8 -50.6 (i.e., each time the ANOSIM p is calculated for a given R, its values will vary slightly). A multidimensional scaling (MDS) plot was construct- ed using XLStat (Addinsoft, New York, NY) based on the Aitchison distance matrix, to illustrate the degree to which the early-stage embryonic and hatched, pre- parturition samples were separated based on their FA compositions. Any differences in the rates at which individual FAs were depleted during embryogenesis were expected to change the overall FA profiles over time; any net change in overall FA profile that occurred between early embryonic and later larval stage samples were revealed in the MDS plot. In order to describe which individual FAs were re- sponsible for the differences in overall FA profiles be- tween early and late stage samples, we calculated the percentages of mass lost ( IML ) for each individual FA: IML = me— x 100%, ( 2 ) ml where me = the average mass of a FA in four samples of early-stage embryos; and ml = the average mass of a FA in four samples of hatched larvae. Comparison to the percentage of total lipid mass lost enabled us to describe which FAs had been depleted most rapidly, and which had been largely conserved. Because the importance of any FA in metabolism may be revealed in a combination of the rates of use and the absolute mass used (i.e., its contribution to the overall loss of lipid), we also described changes in mass of each FA between early and late stage samples and the percentage of total FA mass loss ( TML ) they accounted for: TML = x 100%, (3) TMe - TM, where TMe = the average total mass of all FAs in four samples of early-stage embryos; and TMf - the average total mass of all FAs in four samples of hatched larvae. It is important to note that total lipid masses of samples were independently determined by separate processes from the FA analysis, so total lipid did not simply reflect the summed FA masses. Results Body composition and energy use As they developed, quillback rockfish embryos took on water to gain size while they consumed their stored lipids and, to a lesser extent, protein as energy sources. A typical quillback rockfish embryo gained over one- third its weight in water and lost nearly 20% of its dry mass through the observed course of development, from early embryonic stages to preparturition, hatched larvae (Table 3). Dry mass loss was comprised of 54% protein and 46% lipid. Sewall and Rodgveller: Changes in body composition and fatty acid profile during embryogenesis of Sebastes maliger 213 Figure 2 Protein mass (A) and lipid mass (•) per embryo or larva by developmental stage for quillback rockfish ( Sebastes maliger). Development progresses from left to right: Stage 2 = early embryos (postfertilization); stage 10=hatched larvae (preparturition). Each point repre- sents a single measurement of a composite sample of hundreds of embryos/larvae from one maternal female (n = ll maternal females). Only data from samples for which protein and lipid analyses were both completed are included. For lipid mass, two stage-10 points overlap and are indistinguishable. Lipid: r2 = 0.54, y=-1.90x+48.96. Protein: r2 = 0.35, y=-2.37x+98.02. While lipid and protein were both consumed in sig- nificant amounts, lipid was lost at a greater rate as a proportion of initial lipid mass (34%) than was protein (19%). Though both declined during development, there was greater variability and a weaker correlation be- tween protein mass and developmental stage than lipid mass and stage (Fig. 2). Using these mass losses to estimate energy use (Table 3), a developing embryo consumed a minimum of 0.88 J of energy, on average, with approximately 0.53 J (60%) coming from lipid and 0.35 J (40%) from protein. The slight decrease (8%) in the energy density of dry tissue mass was due to greater proportional losses of lipids than proteins. The 26% decline in total energy content per individual was thus more a reflection of the 20% loss in total dry mass than of the changes in propor- tions of lipid and protein. Oil globule volume The volume of the oil globule in rockfish embryos and larvae reflected both their developmental stage and body composition. OGV declined by 51% from early-stage embryos to hatched larvae. The OGV was highly cor- related with developmental stage (Fig. 3). As embryos progressed through developmental stages, changes in Figure 3 Oil globule volume in nanoliters (nL) by developmen- tal stage for quillback rockfish (Sebastes maliger) embryos and larvae. Development progresses from left to right: Stage 2 = early embryos (postfertilization); stage 10 = hatched larvae (preparturition). Each point represents the mean oil globule volume calculated from diameter measurements of approximately 24 embryos or larvae from each maternal female (n = ll maternal females); two stage-10 points overlap and are indistin- guishable. Only data from samples for which protein and lipid analyses were both completed are included. r2=0.89, y=-1.78x+32.27. OGV indicated trends in overall lipid and protein levels. Simple linear regression analysis indicated that total lipid was significantly dependent upon OGV (Table 4); this held true whether lipid was expressed as lipid mass per individual, or concentration (percentage of wet or dry tissue mass). Protein mass also decreased with OGV, though this relationship was weaker. Fatty add profiles The proportions of fatty acids (FAs) present in quill- back rockfish appeared to change during their early development, as indicated by the significantly differ- ent FA compositions of early embryos versus hatched larvae (ANOSIM R= 0.677, a=0.05, n- 8). An MDS plot of the samples based on their Aitchison matrix distances showed a distinct separation of the early and late-stage FA profiles (Fig. 4). The highest percentage mass losses (> 60%) of indi- vidual FAs were found to occur among the n-11 mono- unsaturated fatty acids (MUFAs) 18:ln-ll, 20:ln-ll and 22:ln-ll; and the polyunsaturated fatty acids (PUFAs) 18:3n-3 (alpha-linolenic acid) and 20:3n-3 (eicosatri- enoic acid) (Eq. 2; Fig. 5). The lowest percentage losses (<20%) occurred for the saturated fatty acid (SFA) 18:0 (stearic acid), the MUFA 24:ln-9 (nervonic acid), and 214 Fishery Bulletin 107(2) Table 4 Simple linear regression parameters relating body composition (response variables) to oil globule volume for quillback rockfish ( Sebastes maliger ) embryos and preparturition larvae. Each sample was a composite of hundreds of larvae from the same parent (n=ll maternal females). Only data from samples for which protein and lipid analyses were both completed were included. Response variable Slope Intercept r2 ANOVA F P Lipid mass 1.13 0.0133 0.672 FJ9=18.45 0.002 Lipid (% wet mass) 247 -0.101 0.943 F7 9=150.14 <0.001 Lipid (% dry mass) 403 20.3 0.701 F;9=99.41 0.001 Protein mass 1.40 0.0537 0.432 F;9 = 6. 85 0.028 the PUFA 20:4n-6 (arachidonic acid). No groups of FAs based on degree of saturation were apparently depleted more rapidly than others, as the percentage mass losses of SFAs, MUFAs, and PUFAs were approximately equiv- alent to the percentage of total FA mass loss (Fig. 5). Some FAs showed relatively little contribution to total FA mass loss despite having large initial masses, indi- cating that they were conserved, particularly the SFA 18:0; and the PUFA 20:4n-6 (Eq. 3, Table 5). Mean- while, the largest absolute mass losses were found for the SFA 16:0 (palmitic acid); the MUFA 18:ln-9 (oleic acid); and the n-3 PUFAs 22:6n-3 (docosahexaenoic acid, DHA) and 20:5n-3 (eicosapentaenoic acid, EPA), which together accounted for 71% of the total loss in FAs. Thus, there were clear differences in the contribu- tions of different FAs to the overall lipid use. | Discussion We found that while both lipid and protein mass are consumed by quillback rockfish embryos during develop- ment, lipid is used more rapidly and contributes a larger portion of total energy than protein. This is consistent with results from other studies of rockfish, and affirms the importance of measuring lipid levels when assess- ing larval condition. However, we also found dif- ferences in the specific rates of use of protein and lipid compared to other rockfish, which illustrates the diversity of patterns of energy use and changes in body composition among species. In our study, OGV was highly correlated with lipid content. This relationship could be important for future studies researching the energetic status of rockfish embryos and preparturition larvae. Using OGV as an indicator of energy reserves at any stage of development, and knowing the rela- tionship between OGV and developmental stage, may allow for interpreting the energetic health of embryos at any developmental stage. This is a considerable advantage for field-based studies, given the difficulty of capturing significant num- bers of gravid females with embryos or larvae at the same developmental stage, and the risks of introducing experimental effects when parents are held until larvae are released. Our results also illustrate that indicators of condition applied to different species should be interpreted with differences in their biochemistries in mind (e.g., in quillback rockfish OGV is strongly related to total lipid, whereas in black rockfish the two are unrelated) (Berkeley et al., 2004). Our study represents the first attempt to char- acterize FA use during embryogenesis for a rock- fish species. Although aquaculture studies have investigated FA requirements for rockfish, these have typically involved manipulating the diets of adults and juveniles (e.g., Lee, 2001) and likely Figure 4 Multidimensional scaling plot of quillback rockfish ( Sebastes maliger ) early-stage embryos (•) and hatched, preparturition larvae (O) according to their fatty acid compositions based on an Aitchison distance matrix. Developmental stage for each sample is given in parentheses. Each sample was a composite of hundreds of larvae from the same parent (n = 8 maternal females). Comparisons only included those samples for which lipid data were available. Sewall and Rodgveller: Changes in body composition and fatty acid profile during embryogenesis of Sebastes maliger 215 Figure 5 Percentage of individual fatty acid (FA) masses lost during embryogenesis of quillback rockfish ( Sebastes maliger) larvae. Percentage mass losses represent differences between average FA masses per larva of four early embryonic samples and four hatched larval samples, in relation to masses present in early embryonic samples. Each sample was a composite of hundreds of larvae from the same parent (n = 8 maternal females). Dashed line indicates difference in total lipid mass between embryos and larvae in relation to lipid mass in embryos. cannot be generalized to developing rockfish embryos and larvae. Although we did not attempt to directly assess the influence of specific FAs on larval survival, our results show FAs are depleted at different rates during embryogenesis. When used in conjunction with data on total lipid levels, the relative abundances of specific conserved FAs of known metabolic importance (e.g., 20:4n-6) may be useful in assessing the condition of embryos and preparturition larvae collected from wild-caught female rockfish. Body composition and energy use Comparisons with other studies of rockfish revealed sub- stantial diversity in the body compositions and energy use patterns of embryos from different Sebastes spe- cies— even after allowing for differences in methods and the high degree of variability in the compositional data. For example, the early stage quillback rockfish embryos studied here had lower lipid (-6.7%) and protein (-14.1%) wet tissue concentrations than those found by Eldridge et al. (2002) for late vitellogenic eggs and early embryos of yellowtail rockfish (S. flavidus) (-12.8% and -21.0%, respectively). Quillback rockfish embryos had lower energy density on a dry mass basis (-24.3 kJ/g) compared with the yellowtail rockfish embryos (-27.1 kJ/g), but because of their larger dry mass, the embryos of quillback rockfish had much greater total energy per individual (3.40 J) than those of the yellowtail rockfish (-1.06 J). The patterns of decline in lipid and protein in quill- back rockfish differed somewhat from those reported by MacFarlane and Norton (1999) for yellowtail rock- fish. They found that lipid as a proportion of wet mass declined 68% and protein decreased by 77%, whereas we found that lipid declined 51% and protein declined 41%. The smaller decreases in lipid and protein con- centration we found may be an artifact of the different ranges of development observed (i.e., our study did not include data from unfertilized oocytes or the earliest stage-1 embryos, when protein and lipid levels were likely higher). The slightly greater decreases in protein concentration than in lipid concentration reported for yellowtail rockfish — opposite to the pattern we found with quillback rockfish — illustrates the high degree of variability among rockfish species. The results of MacFarlane and Norton (1999) for shortbelly rockfish (S. jordani ) followed a pattern similar to ours, with lipid decreasing by 68% and protein by 55%, indicating greater conservation of protein by shortbelly rockfish 216 Fishery Bulletin 107(2) Table 5 Contributions of individual fatty acids (FAs) to total FA mass loss during quillback rockfish ( Sebastes maliger ) embryogenesis based on comparison of average FA masses for four early embryonic (stages 2-3) and four hatched larval (stage 10) samples. Each sample was a composite of hundreds of larvae from the same parent (n = 8 maternal females). Results are ranked by mass loss in nanograms (ng), and grouped by degree of saturation (SFA= saturated fatty acid; MUFA=monounsaturated fatty acid; PUFA=polyunsaturated fatty acid). High variability (low precision), as indicated by coefficients of variation >10%, was found in duplicate samples for 18:lnll (32.1%), 24:ln9 (21.2%) and 24:0 (11.3%). *=trace (<1 ng). Fatty acid Mass (ng) per embryo ±1 SD Mass (ng) per larva ±1 SD Mass loss (ng) % of total FA mass loss SFA 16:0 3250 ±481 2010 ±410 1240 13.5 14:0 660 ±111 300 ±41 360 3.9 18:0 605 ±87 536 ±119 69 0.8 15:0 113 ±9 49 ±13 64 0.7 17:0 73 ±9 37 ±6 36 0.4 20:0 11 ±1 9 ±1 2 <0.1 22:0 * * * <0.1 24:0 * * * <0.1 All SFAs 4710 2940 1770 19.4 MUFA 18:ln9 cis and trans 3450 ±509 2230 ±576 1220 13.4 18:ln7 1260 ±180 731 ±148 529 5.8 16:ln7 1510 ±222 1180 ±340 330 3.6 20:lnll 269 ±100 86 ±44 183 2.0 20:ln9 255 ±42 128 ±16 127 1.4 18:lnll 115 ±35 41 ±53 74 0.8 22:lnll 80 ±36 21 ±10 59 0.7 22:ln9 24 ±5 12 ±1 12 0.1 24:ln9 64 ±9 53 ±13 11 0.1 14:ln5 10 ±2 8 ±3 2 <0.1 All MUFAs 7040 4490 2550 27.9 PUFA 22:6n3 5950 ±959 3850 ±846 2100 23.0 20:5n3 4200 ±832 2240 ±454 1960 21.3 22:5n3 806 ±223 522 ±123 284 3.1 18:2n6 299 ±38 131 ±37 168 1.8 20:4n6 735 ±73 612 ±90 123 1.3 18:3n3 154 ±24 54 ±24 100 1.1 20:3n3 63 ±39 17 ±5 46 0.5 20:2n6 61 ±17 25 ±7 36 0.4 18:3n6 20 ±1 10 ±3 10 0.1 20:3n6 9 ±2 6 ±2 3 <0.1 22:2n6 3 ±1 * * <0.1 All PUFAs 12300 7470 4830 52.7 and quillback rockfish, both of which had lower initial concentrations of protein on a wet mass basis than yel- lowtail rockfish. From a purely energetic perspective, embryos of all three of these rockfish species show a greater decline in energy available as lipid than as protein. The energy density of early-stage quillback rockfish embryos (5.24 J/mg) was similar to the typical value for marine spawning species of 6.0 J/mg reported by Kamler (1992). Changes in the energy density of wet tissue mass were largely a reflection of changes in the percent moisture; whereas changes in the dry tissue composition contributed less. Energy density on a dry mass basis was similar to the value for fish eggs of 23.48 J/mg reported by Wootton (1979) as an average across many species. This is not surprising, given that interspecific variation in the energy density of fish eggs is relatively low (Kamler, 1992), compared with the range of egg sizes and total energy contents. The distinction between viviparous and ovoviviparous is a consideration in interpreting mass loss and energy data in our study because it hinges on whether the em- Sewall and Rodgveller: Changes in body composition and fatty acid profile during embryogenesis of Sebastes maliger 217 bryos developing inside the mothers’ bodies are supplied with maternal nutrients (viviparous), or rely entirely on their yolk sacs (ovoviviparous). Quillback rockfish have been described as viviparous (MacFarlane and Norton, 1999), and ovoviviparous (Matala et al., 2004). Previ- ous research using radiocarbon-labeled amino acids found that embryos of black rockfish (S. melanops) took up nutrients from intraovarian fluid, but only at very late stages of development — presumably after they had hatched and their mouths and digestive systems were sufficiently functional (Yoklavich and Boehlert, 1991). MacFarlane and Bowers (1995) also found evidence of matrotrophy (postfertilization maternal nutrient pro- visioning) occurring in yellowtail rockfish because a radio-labeled phospholipid was transferred from moth- ers to embryos before their mouths opened, and the amount increased as they developed. Reviews of these and other studies have thus supported viviparity in rockfish (e.g., Parker et al., 2000). The reduction in dry tissue mass seen among the quillback rockfish embryos in our study was lower than the 25% to 55% range of dry mass losses typically seen in strict lecithotropes (MacFarlane and Bowers, 1995), which rely entirely on nutrients provided to the egg before fertilization, sug- gesting that quillback rockfish are also partly matrotro- phic. The degree to which nutrition is obtained from the yolk rather than from maternal intraovarian fluids is unclear for quillback rockfish; therefore it is important to view data regarding mass loss and energy use given here as minimums. It is likely that maternal traits (e.g., the size and age of the female parent) influence the biochemical compositions of rockfish embryos and larvae (Sogard et al., 2008). This introduces the possibility of maternal effects confounding the relationship between develop- mental stage and body composition (e.g., if our samples were biased towards larger females yielding the earlier stages of embryos). However, it is likely that develop- mental processes accounted for most of the differences that we found between early-stage embryos and hatched larvae. Developmental stage showed a much stronger relationship to lipid concentration (r2 = 0.87) than did maternal length (r2=0.39). Maternal length was only weakly correlated with developmental stage (r2=0.26), and this correlation was largely driven by the pres- ence of one large fish with stage 10 larvae. Removing this fish and its larvae resulted in virtually no rela- tionship between maternal length and developmental stage (r2 = .15). This highlights one of the conclusions that can be drawn from our data: developmental stage should be accounted for when investigating maternal effects among wild caught fish with progeny at various stages. If quillback rockfish preferentially use lipid as an energy source over protein, it would be useful to inves- tigate how various maternal traits influence the relative rates of lipid and protein loss in embryos. For example, do embryos from older, larger, or fatter parents have greater lipid reserves, and do they exhibit lower rates of protein loss? Why not simply use size or total energy content as indicators of viability? Such an approach is indicated by our finding that changes in total energy content per larva largely reflected changes in dry mass from early to late stages, rather than changes in the proportions of lipid and protein. In addition, there is great vari- ability among species in the size and energy content of eggs and embryos — the early stage quillback rockfish embryos in our study were on average more than 2.6 times heavier on a dry mass basis than yellowtail rock- fish embryos (Eldridge et al., 2002). Greater larval size may also confer advantages through reduced predation and increased range of feeding opportunities, and was probably the force driving the uptake of water during early development that we observed. However, various studies have found no relationship between egg size and offspring viability (reviewed in Kamler, 1992). Straight- forward interpretation of the relationship of egg or em- bryo size and total energy content to larval viability is confounded by findings suggesting that larvae from smaller eggs often use yolk energy for growth more ef- ficiently than those from larger eggs, and may undergo compensatory growth in later development (reviewed in Kamler, 1992). Even under conditions of food scarcity, where larger larvae may be expected to be at an advan- tage, results have been inconsistent; for example, larval length did not appear related to starvation resistance of black rockfish larvae (Berkeley et al., 2004). Oil globule volume Given the importance of lipid as an energy source for developing quillback rockfish embryos, the strong cor- relation of OGV with total lipid we found suggests that OGV may serve as an indicator of energetic status. Some maternal trait, such as age (e.g., Berkeley et al., 2004), may strongly influence OGV and be responsible for the variability. Investigating changes in the lipid class com- ponents (e.g., TAG and polar lipids) of the oil globules, as well as whole embryos, could provide information useful for better understanding the relationship of the oil globules to condition. The strength of the relationship between OGV and larval survival should also be inves- tigated experimentally with quillback rockfish larvae. Using OGV as an indicator of energetic status represents a potentially large savings in resources required, com- pared with analytical chemistry techniques. Fatty acid profile The major FA components of the lipids in quillback rockfish embryos and larvae were generally similar to those reported elsewhere for many species of adult fish (reviewed in Tocher, 2003): predominantly the n-3 PUFAs 22:6n-3 and 20:5n-3; 20:4n-6 as the main n-6 PUFA; large quantities of the MUFA 18:ln-9; and 16:0 and 18:0 as the main SFAs. Previous researchers have also reported high levels of n-3 PUFAs in marine fish eggs (e.g., Tocher & Sargent, 1984); however, there can be marked interspecific differences in the precise order 218 Fishery Bulletin 107(2) of FA abundances. For example, in contrast to the quill- back rockfish embryos studied here, which showed the n-3 PUFAs 22:6n-3 and 20:5n-3 in greatest abundance, Tveiten et al. (2004) reported that of 16 FAs they inves- tigated in embryos of the spotted wolffish ( Anarhichas minor), 18:ln-9 was predominant. Caution must be used when attempting to apply condition indices based on FA amounts or proportions derived from other species. As lipids are broken down for use during development, the resulting FAs may be conserved as structural components of new tissues or metabolic compounds, modified into new FAs, or con- sumed as energy sources, and the timing and extent to which specific FAs are used varies considerably among species (reviewed in Tocher, 2003). In some marine fish, FAs appear to be utilized in a non-selec- tive fashion (e.g., in order of their abundance) while in others, some FAs have been preferentially retained. For example, retention of 20:4n-6 was found to occur in Murray cod ( Maccullochella peelii peelii ) and trout cod (Maccullochella macquariensis ; Gunasekera et al., 1999), Senegalese sole (Solea senegalensis; Mourente and Vazquez, 1996), and spotted wolffish ( Anarhi- chas minor ; Tveiten et al., 2004); this PUFA was also used less rapidly than the total lipid for the rockfish embryos in this study. Greater retention of 20:5n-3 has been reported in Atlantic halibut (Hippoglossus hippoglossus ; Ronnestad et al., 1995), but this did not occur for quillback rockfish here. Tveiten et al. (2004) found that spotted wolffish embryos had lower ratios of 20:4n-6 to 20:5n-3 than those generally deemed necessary for survival in other species. In spotted wolf- fish embryos, the proportion of 16:0 increased, while 18:ln-9 decreased (Tveiten et al., 2004); for quillback rockfish, these FAs were used at the same rate as total FAs. This suggests that species differences must be considered in any assessment of the FA composition of developing fish. Saturated fatty acids and monounsaturated fatty acids The SFAs 16:0 and 18:0, and MUFAs that can be derived from them (e.g., 16:ln-7, 18:ln-9), are unlikely candidates for use as indicators of quillback rockfish nutritional or energetic status due to their relatively high abundances and the ability of all organisms to biosynthesize them. Any deficiencies in these FAs could be readily inferred from low total lipid levels. The MUFAs with high percentage mass losses were generally present in very low amounts and likely were not of high metabolic importance. For example, the MUFA 22:ln-ll, which is likely derived from calanoid copepods and transferred up through higher trophic levels in marine food chains (Saito and Kotani, 2000), was found to have the greatest rate of decrease in mass during larval development. However, its small initial mass and general absence from structural lipids in fish (Tocher, 2003) makes it likely to serve only as a minor energy source for developing quillback rockfish embryos. Polyunsaturated fatty acids The finding that 20:4n-6, which was the most abundant n-6 PUFA, was largely conserved seems consistent with its role as an important metabolic end product rather than a general energy source. As a precursor to the eicosanoids, a physiologically active and diverse group of hormone-like compounds, 20:4n-6 is believed to play a significant role in a variety of functions, including inducement of spawning, intercellular signaling, stress tolerance, immune response, inflammatory response, blood clotting, and is likely essential to normal growth and development (reviewed in Bell & Sargent, 2003; Tocher 2003). Several aquaculture studies have indi- cated that supplementing broodstock diets with 20:4n-6, within optimal concentration ranges or ratios to other FAs, results in improved egg and larval quality for a variety of marine fish species (reviewed in Bell & Sar- gent, 2003). Many marine fish seem to need 20:4n-6 in their diets and are unable to manufacture it from pre- j cursors (Mourente and Tocher, 1993; reviewed in Bell & Sargent, 2003); the levels of 20:4n-6 in embryos there- fore likely reflect the quality of maternal provisioning. While measuring the relative abundance of 20:4n-6 may be useful in assessing condition of quillback rockfish embryos, further investigation is needed to determine what levels of 20:4n-6 may be considered deficient, and what specific effects may arise from that deficiency. The PUFAs 20:5n-3 and 22:6n-3 were the most abun- dant FAs measured in quillback rockfish embryos and preparturition larvae, and decreased at approximately the same rate as total lipids. While 20:5n-3 and 22:6n- 3 are important metabolic end products, they can also be consumed as major energy sources during the early life history of many marine fishes (reviewed in Toch- er, 2003). Due to their abundance and role as energy sources, the amounts of 20:5n-3 and 22:6n-3 in quill- back rockfish embryos are reflected in total lipid levels, and would not be informative as additional indicators of condition. The remaining PUFAs were used more quickly than the total lipid, were generally not very abundant, and are likely of limited importance. For example, 18:3n-3, an essential fatty acid derived from marine plants, can serve as a precursor to both 20:5n-3 and 22:6n-3 in some organisms, following a metabolic pathway that is similar across widely varying taxa (reviewed in Tocher, 2003). However, the high abundances of 20:5n-3 and 22:6n-3, in combination with the relatively low levels of 18:3n-3 (<1% total FA mass), suggest that they were being synthesized from 18:3n-3 to supplement the ma- ternally-provisioned amounts, it was only to a minor degree. Other research suggests that marine fish are largely incapable of synthesizing 20:5n-3 and 22:6n- 3 from 18:3n-3 and they obtain these essential fatty acids from their diets (Tocher, 2003), in which case the embryos are likely using 18:3n-3 as a relatively small energy source. Similarly, 18:2n-6 may be of limited importance, as marine fish have limited ability to con- vert it to the metabolically-important PUFA 20:4n-6, Sewall and Rodgveller: Changes in body composition and fatty acid profile during embryogenesis of Sebastes maliger 219 and it was present at relatively low levels in quillback rockfish embryos. Although both lipid and protein are consumed during quillback rockfish embryogenesis, lipid is used more rapidly and contributes a greater portion of the total energy expended. Lipid is typically the most variable dry mass component of fish eggs, showing significant differences between populations and within a popula- tion over time; additionally, lipid concentration has been used as an indicator of larval viability for several spe- cies (reviewed in Kamler, 1992). Given the importance of lipid as an energy source, the strong relationship be- tween OGV and lipid levels confirms the utility of OGV as an indicator of differences in condition of quillback rockfish embryos and preparturition larvae. Differences in FA profiles of early embryos and preparturition lar- vae indicate FAs are depleted at different rates during embryogenesis. More rapidly used FAs may contribute more to lipid energy use or serve as precursors in the synthesis of other FAs, while conserved FAs likely are incorporated into tissues or hormone-like compounds. The conservation of 20:4n-6, the most abundant n-6 PUFA, indicates that this essential fatty acid may well reflect the quality of maternal provisioning. The high degree of interspecific variability in body composition and energy use patterns among rockfish illustrates the need for data gathered from the species of interest, in order to make the most accurate models of energy use and most appropriate indicators of condition. Acknowledgments We thank R. Heintz for advice on study methodology and data analysis, M. Larsen for generating the fatty acid data, and R. Bradshaw for generating the protein, mois- ture, and ash content data. We also thank J. Maselko for developing the computer code for creating the Aitchison distance matrix and ANOSIM statistical analysis, C. Lunsford for assisting with field sampling, and J. Heifetz and L. Schaufler for providing helpful commentary and edits on earlier drafts. Literature cited Aitchison, J. 1992. On criteria for measures of compositional difference. Math. Geol. 24:365-379. Bell, J. G., and J. R. Sargent. 2003. Arachidonic acid in aquaculture feeds: current status and future opportunities. Aquacult. 218:491- 499. Berkeley, S. A., C. Chapman, and S. M. Sogard. 2004. Maternal age as a determinant of larval growth and survival in a marine fish, Sebastes melanops. Ecology 85(5):1258-1264. Brett, J. R. 1995. Energetics. In Physiological ecology of Pacific salmon (Groot, C., Margolis, L., and W.C. Clarke, eds.), p. 1-68. UBC Press, Vancouver, Canada. Christie, W. W. 2003. Lipid analysis: Isolation, separation, identification and structural analysis of lipids, 416 p. Oily Press, Bridgwater, U.K. Clarke, K. R., and R. M. Warwick. 1994. Change in marine communities: An approach to statistical analysis and interpretation, 144p. Plymouth Marine Laboratory, Plymouth, U.K. Craig, J. F., M. J. Kenley, and J. F. Tailing. 1978. Comparative estimations of the energy content of fish tissue from bomb calorimetry, wet oxidation and proximate analysis. Freshw. Biol. 8:585-590. Eldridge, M. B., E. C. Norton, B. M. Jarvis, and R. B. MacFarlane. 2002. Energetics of early development in the viviparous yellowtail rockfish. J. Fish Biol. 61:1122-1134. Gunasekera, R. M., S. S. De Silva, and B. A. Ingram. 1999. Early ontogeny-related changes of the fatty acid composition in the percichthyid fishes trout cod, Maccullo- chella macquariensis and Murray cod, Maccullochella pelii peelii. Aquat. Living Resour. 12:219-227. Horwitz, W., ed. 2002. Official methods of analysis, 17th ed. AOAC (Association of Analytical Communities) International, Gaithersburg, MD. Kamler, E. 1992. Early life history of fish, an energetics approach. 267 p. Chapman and Hall, London. Lee, S. M. 2001. Review of the lipid and essential fatty acid require- ments of rockfish ( Sebastes schlegeli). Aquacult. Res. 32 ( s 1 ) : 8— 17. MacFarlane, R. B., and M. J. Bowers. 1995. Matrotrophic viviparity in the yellowtail rockfish Sebastes flavidus. J. Exp. Biol. 198:1197-206. MacFarlane, R. B., and E. C. Norton. 1999. Nutritional dynamics during embryonic develop- ment in the viviparous genus Sebastes and their applica- tion to the assessment of reproductive success. Fish. Bull. 97:273-281. Matala, A. P., A. K. Gray, J. Heifetz, and A. J. Gharrett. 2004. Population structure of Alaskan shortraker rock- fish, Sebastes borealis, inferred from microsatellite variation. Environ. Biol. Fishes 69:201-210. Mourent.e, G., and D. R. Tocher. 1993. Incorporation and metabolism of 14C-labeled poly- unsaturated fatty acids in wild-caught juveniles of golden grey mullet, Liza aurata, in vivo. Fish Physiol Bio- chem. 12(2):119— 130. Mourente, G., and R. Vazquez. 1996. Changes in the content of total lipid, lipid classes and fatty acids of developing eggs and unfed larvae of the Senegal sole, Solea senegalensis Kaup. Fish Physiol. Biochem. 15(3):221-235. Norton, E. C., R. B. MacFarlane, and M. S. Mohr. 2001. Lipid class dynamics during development in early life stages of shortbelly rockfish and their application to condition assessment. J. Fish. Biol. 58:1010—1024. Parker, S. J., S. A. Berkeley, J. T. Golden, D. R. Gunderson, J. Heifetz, M. A. Hixon, R. Larson, B. M. Leaman, M. S. Love, J. A. Musick, V. M. O’Connell, S. Ralston, H. J. Weeks, and M. M. Yoklavich. 2000. Management of Pacific rockfish. Fisheries 25(3): 22-30. Ralston, S., and D. F. Howard 1995. On the development of year-class strength and cohort 220 Fishery Bulletin 107(2) variability in two northern California rockfishes. Fish. Bull. 93:710-720. Ronnestad, I., R. N. Finn, I. Lein, and O. Lie. 1995. Compartmental changes in the contents of total lipid, lipid classes and their associated fatty acids in developing yolk-sac larvae of Atlantic halibut, Hip- poglossus hippoglossus ( L . ) . Aquacult. Nutr. 1:119— 130. Saito, H. and Y. Kotani. 2000. Lipids of four boreal species of calanoid copepods: origin of monoene fats of marine animals at higher trophic levels in the grazing food chain in the subarctic ocean ecosystem. Mar. Chem. 7:69-82. Sogard, M. S., S. A. Berkeley, and R. Fisher. 2008. Maternal effects in rockfishes Sebastes spp.: a comparison among species. Mar. Ecol. Prog. Ser. 360:227-236. Tocher, D. R. 2003. Metabolism and functions of lipids and fatty acids in teleost fish. Rev. Fish. Sci. 11(2):107-184. Tocher, D. R., and J. R. Sargent 1984. Analyses of lipids and fatty acids in rope roes of some Northwest European marine fish. Lipids 19(71:492-499. Tveiten, H., M. Jobling, and I. Andreassen. 2004. Influence of egg lipids and fatty acids on egg viabil- ity, and their utilization during embryonic development of spotted wolf-fish, Anarhichas minor Olafsen. Aquacult. Res. 35(21:152-161. Watanabe, T. 1982. Lipid nutrition in fish. Comp. Biochem. Physiol. B 73(11:3-15. Wootton, R. J. 1979. Energy costs of egg production and environmental determinants of fecundity in teleost fishes. Symp. Zool. Soc. Lond. 44:133-159. Yamada, J., and M. Kusakari. 1991. Staging and the time course of embryonic devel- opment in kurosai, Sebastes schlegeli. Environ. Biol. Fishes 30:103-110. Yoklavich, M., and G. W. Boehlert. 1991. Uptake and utilization of 14C-glycine by embryos of Sebastes melanops. Environ. Biol. Fishes 30:147- 153. 221 Abstract— Depth data from archival tags on northern rock sole (Lepidop- setta polyxystra ) were examined to assess whether fish used tidal cur- rents to aid horizontal migration. Two northern rock sole, out of 115 released with archival tags in the eastern Bering Sea, were recovered 314 and 667 days after release. Both fish made periodic excursions away from the bottom during mostly night- time hours, but also during particu- lar phases of the tide cycle. One fish that was captured and released in an area of rotary currents made vertical excursions that were cor- related with tidal current direc- tion. To test the hypothesis that the fish made vertical excursions to use tidal currents to aid migra- tion, a hypothetical migratory path was calculated using a tide model to predict the current direction and speed during periods when the fish was off the bottom. This migration included limited movements from July through December, followed by a 200-km southern migration from January through February, then a return northward in March and April. The successful application of tidal current information to pre- dict a horizontal migratory path not only provides evidence of selective tidal stream transport but indicates that vertical excursions were con- ducted primarily to assist horizontal migration. Manuscript submitted 29 July 2008. Manuscript accepted 19 December 2008. Fish. Bull. 107:221-234 (2009). The views and opinions expressed or implied in this article are those of the author and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Evidence of the selection of tidal streams by northern rock sole ( Lepidopsetta polyxystra ) for transport in the eastern Bering Sea Daniel G. Nichol (contact author) David A. Somerton E-mail address for contact author: dan.nichol@noaa.gov Alaska Fisheries Science Center National Marine Fisheries Service, NOAA 7600 Sand Point Way NE Seattle, Washington 98115 Northern rock sole (Lepidopsetta polyxystra ) in the eastern Bering Sea off Alaska reportedly migrate from summer feeding grounds to deeper spawning grounds in winter (Shub- nikov and Lisovenko, 1964; Fadeev, 1965). Although migration routes are poorly understood, at least some indi- viduals are thought to migrate long distances between summer and winter grounds. Shubnikov and Lisovenko (1964) suggested that some northern rock sole migrate from Unimak Island in the Aleutian Islands to areas north- east of the Pribilof Islands between April and July, covering a distance of more than 500 km. As a means to understand better how this migration occurs, we focus here on one poten- tial mechanism that northern rock sole use, that is, the opportunistic or exclusive use of selective tidal stream transport. Selective tidal stream transport is a mechanism by which aquatic animals can assist their horizontal migration by actively changing their vertical position in the water column, timed to coincide with tidal currents flow- ing in a preferred direction. Selec- tive tidal stream transport has been documented for a variety of aquatic animals (Forward and Tankersley, 2001; Gibson, 2003) and has been extensively documented in the North Sea for European plaice (Pleuronectes platessa) (Kuipers, 1973; Rijnsdorp and van Stralen, 1985; Metcalfe et al., 1990; Fox et al., 2006; Metcalfe et al., 2006). Even before the use of electronic fish tags, it was recognized that some flatfish species selectively leave the bottom during periods of a preferred tidal current direction (De Veen, 1967; Harden Jones et al., 1979). More recent work with archi- val tag data, used in combination with tide data (e.g., Hunter et ah, 2004b), has highlighted the impor- tance of both tide and diurnal fac- tors in flatfish migration. For plaice and other flatfish species, the vertical movements (i.e., selective tidal stream transport) vary diurnally; most excur- sions are made away from bottom dur- ing the night (De Veen, 1967; Cadrin and Westwood, 2004; Hunter et ah, 2004b; Walsh and Morgan, 2004). Juvenile flatfishes (Champalbert et ah, 1992; Burrows, 1994), including northern rock sole (Hurst and Duffy, 2005), also show some preference for nighttime activity. When flatfishes use selective tidal stream transport, the timing of verti- cal excursions away from the bottom can be combined with predictions of tidal current velocity to construct hypothetical migration trajectories (Arnold and Holford, 1995). For this estimation to be successful, verti- cal excursions need to be accurately identified and all directed horizontal movement must be restricted to those off-bottom periods. For species known to use tidal currents, selective tidal stream transport may not be used in all habitats because the strength or direction of the currents may be un- suitable for migration. For example, when European plaice inhabit areas of weak tidal currents in the North Sea, they migrate horizontally, stay- ing near the seafloor where there is 222 Fishery Bulletin 107(2) 56°N- 54°N- Fish Release Recovery Northern O O Southern O O PMEL Mooring -60° N ■58°N -56°N — 168°W 1 164°W 100 km 1 160°W 168°W 172°W Alaska 164°W 160°W I Figure 1 Map of release and capture locations for two female northern rock sole (Lepidopsettci polyxystra) tagged with archival tags in the eastern Bering Sea off Alaska in 2003. The locations of the Pacific Marine Environmental Laboratory (PMEL) subsurface moorings where tidal current velocity data (speed and direction) were collected are also shown. Gray lines with numbers indicate the bathymetric contours (in m) in the area. no tidal assistance (Hunter et al., 2004a; Metcalfe et al., 2006). In addition to accurately determining peri- ods when a fish is off-bottom, successful selective tidal stream transport modeling also requires the ability to accurately predict tidal currents. Tidal currents over most of the eastern Bering Sea shelf have characteristics that offer fish a mechanism that assists horizontal movement. Such currents range from rotary motion on the central continental shelf (e.g., east of Pribilof Islands) to bidirectional motion along the Alaska Peninsula (Kowalik, 1999; Pearson et al., 1981), and a strong semidiurnal component everywhere which offers fish two opportunities within a 24-h period to use currents to migrate in a particular direction. Near-bottom tidal current velocities over much of the eastern Bering Sea shelf average approximately 20 cm/s and peak to over 50 cm/s, which if used for transport, could provide a significant increase in migration veloc- ity. Moreover, the rotary nature of the currents offers an intriguing mechanism for transport because it can provide a means of transport in any direction a fish chooses. Here, we examine depth and time data from archival tags attached to two northern rock sole to determine whether vertical excursions are related to diel and tidal influences, and whether a simple model of selective tidal stream transport can be used to construct a hypotheti- cal horizontal migration path that is consistent with the observed tag release and recovery locations Methods Tagging Two northern rock sole were recovered from among 115 released with attached electronic data storage tags in the eastern Bering Sea between 4 June and 26 July 2003. Release locations were approximately 200 km northeast of St. Paul Island in the Pribilof Islands (northern fish) and 18 km northwest of Unimak Island (southern fish) (Fig. 1). Fish were initially captured with a bottom trawl, tagged, and released during the course of the annual eastern Bering Sea bottom trawl survey (Acuna and Nichol and Somerton: Tidal stream transport of Lepidopsetta polyxystra in the eastern Bering Sea 223 Lauth, 2008). The two recovered fish, both captured by commercial trawlers, were a 34-cm-total-length (TL) (at release) female at liberty for 314 days (northern fish) and a 40-cm-TL (at release) female at liberty for 667 days (southern fish). Both fish were assumed to be mature because they were larger than the reported mean size at maturity of 32.8 cm (Stark and Somerton, 2002). The fish were tagged with Lotek wireless LTD-1100 (St. John’s, NF, Canada) data storage tags. Tags were attached to the eyed-side, just below the anterior end of the dorsal fin with a 0.5-mm diameter stainless- steel wire. The wire was inserted through two points on the tag, through the epaxial musculature above the pectoral fin, and affixed on the blind side of the fish by using oval plastic backing. The two wire ends were fastened on the outside of the backing with a crimped connector sleeve. Tag data, including depth (pressure) and tempera- ture, were recorded at 0.5-h or 1-h time intervals, total- ing 12,015 and 16,346 data pairs for the northern and southern fish, respectively. Two sampling intervals were used because, as a memory management function of the tags, the frequency of recordings decreased with the time at liberty. Depth had a resolution of 0.58 m when fish remained at depths less than 150 m, and 1.2 m if the fish exceeded 150 m; temperature had an accuracy of ±0.3°C. The northern tag recorded for the entire 314 days the fish was at liberty, whereas the southern tag recorded for 620 of 667 days at liberty before the bat- tery died. Tide prediction Tidal height and current speed and direction were esti- mated at the midpoint location between fish release and recovery (northern fish: 58°18'N, 167°02'W; southern fish: 54°55'N, 164°31'W) for each depth measurement using the OTIS Tidal Inversion Software (Oregon State University, Corvallis, OR) which was created with solu- tions specifically for the eastern Bering Sea (Egbert et al., 1994; Egbert and Erofeeva, 2002). To test the accuracy of the tide model, speed and direction were also estimated at the site of an oceanographic moor- ing maintained by the Pacific Marine Environmental Laboratory (NOAA, Seattle, WA) near each fish (Fig. 1) and compared to the measured bottom current data. The northern mooring was located approximately 118 km west of the northern fish location and collected data from October 2004 to April 2005; the southern mooring was located approximately 66 km north of the southern fish location and collected data from March 1995 to Sep- tember 1995. Each data set consisted of hourly current velocity vectors (u=east-west component, v=north-south component) over a period of 193 days. Identification of vertical excursions Time intervals during which the fish were off bottom were identified as follows. Measured tag depths were first corrected for tide height variation by subtracting the tide height predicted by the tide model. Because along-slope movements can be confused with off-bottom movements, the difference in bathymetric complexity between northern and southern tag release locations dictated differences in the subsequent analysis. For the northern fish, distance off bottom was calcu- lated as bottom depth minus tag depth where bottom depth was estimated in two stages. First, daily bottom depth was chosen as the maximum tag depth during each 24-h period, on the assumption that northern rock sole contact the bottom at least once a day. Second, bot- tom depths for each 0.5-h or 1-h recording were estimat- ed by linearly interpolating between the times of daily maxima (proc Expand; SAS, vers. 8.02, SAS Inst., Inc., Cary, NC). Times of vertical excursions were identified as those when the off-bottom distance exceeded 2 m. Discrete excursions away from the bottom were defined as groups of successive off-bottom time recordings. For the southern fish, which resided in steeper, more rugged terrain, distance off bottom was calculated similarly by using 6-h rather than 24-h time windows. In addition, during times when horizontal movements appeared to be occurring in steep terrain, off-bottom distance was calculated by using estimates of bottom depth for still more frequent intervals, assuming that bottom depth was identical to tag depth during each tag recording if tidal fluctuations were obvious in the original tag depth data (i.e., not corrected for tide). The assumption is that the fish was on bottom when tidal fluctuations were recognized. The southern fish was con- sidered off bottom when off-bottom distances exceeded 3 m. Following the criteria used for the northern fish, each 0.5-h or 1-h depth value was designated as either on or off bottom, and discrete excursions were identi- fied. Analyses of excursions for the southern fish were limited to summary statistics because of the difficulty in accurately identifying off-bottom periods (see Discus- sion section). Diel and tidal influence on vertical excursions To determine whether the likelihood of excursions dif- fered between day and night, each 0.5-h or 1-h record collected by a tag was first designated as daytime or nighttime based on the predicted times of sunrise and sunset at the midpoint location between fish release and recovery. Daily sunrise and sunset times were calculated by using an algorithm obtained from the U.S. Naval Observatory, Astronomical Applications Department (Washington, DC). The percentage of off-bottom time recordings occurring during the day and night were then calculated for each fish, and the timing and duration of excursions were plotted with respect to day and night. To determine if vertical excursions of the northern fish were selectively made with respect to tidal cur- rent direction, patterns in tidal current direction were examined using compass plots (function Compass (x,y); Matlab, vers. 7.5.0.342, The MathWorks, Inc., Natick, MA); plots of current direction during hours when the fish was off bottom were compared with plots of all 224 Fishery Bulletin 107(2) the available current directions. In addition, the sig- nificance of current speed and direction in determining whether the northern fish was off bottom was tested using Generalized Additive Modeling (GAM; Venables and Ripley, 1994). This was done by modeling off-bot- tom status (coded 0 for on bottom and 1 for off bottom) as a binomial response to a smooth function of current speed and direction, with significance based on analy- sis of deviance. The test was conducted independently for each month that the fish was at liberty, excluding daytime hours when fish remained on the bottom (see Results section). Selective tidal transport model To determine if the selective use of tidal currents by northern rock sole was an important component of their seasonal horizontal migrations, we developed a selec- tive tidal stream transport model similar in its basic design to that discussed in Arnold and Holford (1995). Assuming that all horizontal movement occurred during off-bottom periods, we constructed a migration path as follows. Starting from the release location, the fish was assumed to drift at the current speed predicted for each off-bottom time and in the mean current direction during each vertical excursion. In addition, the fish was allowed to swim at a specified speed, also in the mean current direction during each vertical excursion. The specified fish swimming speed was held constant over the entire path, but because the true swimming speed was unknown, this speed was iteratively varied until the distance between the final path location and the actual capture location was minimized. Thus, for each 0.5-h or 1-h recording time starting with the release location, latitude and longitude positions were advanced by using the combined drift and swimming speed and mean current direction for each excursion. Tag locations along the selective tidal stream transport path were converted to latitude and longitude positions using great circle formulae. To verify the accuracy of the predicted migration path, the predicted depth at each location along the path was plotted against the bottom depth (maximum 24-h depth) measured by the tag. Path depths were pre- dicted with inverse distance-weighted surface interpola- tion (ArcMap 9.2 with Spatial Analyst Extension, ESRI, Redlands, CA) using National Imagery and Mapping Agency (NIMA) depth sounding data for the eastern Bering Sea continental shelf. Results Depth data and vertical excursions The archival tag depth data contained relatively high frequency variation from three sources. First, verti- cal excursions away from the bottom were identifiable by sharp decreases in tag depth (Fig. 2). Second, tidal height fluctuations were evident in the tag depth record (Fig. 2 A inset), an indication that fish were settled on the bottom. Third, rapid changes in depth resulted from horizontal movements along steep bottom gradients. The identification of vertical excursions was clearer for the northern fish than for the southern fish. The northern fish inhabited areas of relatively flat bottom where estimated tag bottom depths ranged from 48 to 74 m; therefore rapid depth changes clearly reflected vertical excursions away from the bottom. The south- ern fish, by comparison, inhabited an area of complex bathymetric contours where bottom depths collected from tags ranged from 9 to 161 m. Movements along a steep bottom gradient were evident for the southern fish, particularly during April 2004 when fish depth increased and decreased more than 100 m within a 24-h period (Fig. 2B). Because of this complexity, there were occasions when it was unclear whether changes in tag depth resulted from a vertical excursion or a quick movement along a bottom gradient. For this reason, not all vertical excursions were identifiable for this fish. More gradual movements across bottom gradients were observed for both fish, and most often coincided with periods of vertical excursions, indicating that the excursions were related to the horizontal movement of the fish. In a few cases with the southern fish, however, movements across bottom gradients occurred in the absence of vertical excursions (Fig. 2B inset). The frequency, duration, and distance of vertical ex- cursions away from the bottom were similar for the two fish. A total of 78 distinct excursions away from the bottom were identified for the northern fish during a period of 314 days, and 154 excursions were identified for the southern fish during a period of 620 days (Table 1). These excursions were relatively rare, accounting from 2.0% (southern fish) to 2.6% (northern fish) of the time at liberty. Average excursion durations were 2.6 hours (northern fish) and 2.1 hours (southern fish); average excursion extent (i.e., maximum distance off bottom) was 14 m with a maximum of 64 m (Table 1). The frequency of vertical excursions varied seasonally; most excursions occurred from winter to spring for the northern fish and during fall and spring for the south- ern fish. Excursions were infrequent during summer months (Fig. 2). The timing of vertical excursions was related to both diel and tidal factors. For the northern fish, 90% of the excursions occurred at night, whereas for the southern fish, 85% occurred at night (although not all vertical excursions could be identified with certainty). Both fish underwent vertical excursions that sometimes occurred over a series of consecutive nights (Fig. 3). In addition to being limited to nighttime, vertical excursions oc- curred during particular stages of the tide cycle. For example, during the beginning of September 2003, the southern fish made consecutive nightly excursions, but only before the dominant low tide (Fig. 3B). Examina- tion of tidal current directions (northern fish only), revealed the fish did not indiscriminately choose night- time periods, but made nightly vertical excursions only when tidal currents were in certain directions. For ex- Nichol and Somerton: Tidal stream transport of Lepidopsetta polyxystra In the eastern Bering Sea 225 ample in January, when the northern fish made vertical excursions with greatest frequency, it did so when tidal currents were southerly directed, yet the prevailing nighttime tidal currents were directed toward the west and northeast (Fig. 4). During months with frequent excursions, the probability of being off-bottom was not significantly related to current speed but was highly significantly related to current direction (Table 2; GAM (generalized additive modeling) test). Considering the rotary nature of the tidal current where the northern fish resided, the timing of vertical excursion was selec- tive, as opposed to random, with respect to tidal current direction. This selection was particularly evident during January when nighttime periods were sufficiently long to allow for two separate southerly directed currents in a single night. Coincident to these duel nightly southern currents, the fish sometimes made two separate nightly vertical excursions (Fig. 3A). Migration path The predicted migration path based on selective tidal stream transport for the northern fish extended for 503 km and ended 0.26 km from the reported capture 226 Fishery Bulletin 107(2) Table 1 Excursion duration and distance away from the bottom for two northern rock sole ( Lepidopsetta polyxystra) released with archi- val tags and recaptured in the eastern Bering Sea between 2003 and 2005. Minimum durations were limited to the collection frequencies of the tags which were 0.5 hour and 1.0 hour, respectively, for the northern and southern fish. Mean distances were weighted by the sampling frequency for the tags. Numbers in parentheses indicate the average maximum distance off bottom among excursions. Not all excursions could be identified with certainty for the southern fish because of the variable bathymetric terrain in the area where the southern fish resided. Duration away from bottom (h) Distance away from bottom (m) Identified Time Min. Max. Mean Min. Max. Mean excursions recordings Northern fish 0.5 6.5 2.6 2.2 54.2 10.4(13.8) 78 316 Southern fish 1.0 7.0 2.1 3.0 63.7 12.2(13.4) 154 332 Table 2 Significance of tidal current direction and speed on the probability of a northern rock sole ( Lepidopsetta polyxystra) being off the bottom. Data are from one fish tagged in the more northerly area of the eastern Bering Sea shelf. Generalized additive modeling (GAM) with a binomial error term (0=on bottom, l=off bottom) was used to test probabilities for each month, during nighttime. n = the total number of timed tag depth recordings per month. Number in parentheses indicates the number of recordings when the fish was away from the bottom. Months of fewer than 10 nighttime off-bottom observations are excluded. Month Chi-square P-value n Direction Speed Direction Speed Oct. 03 15.0 1.3 0.0018 0.7310 855 (28) Nov. 03 9.2 8.6 0.0236 0.0329 969(12) Dec. 03 10.9 2.7 0.0106 0.4057 1079(12) Jan. 04 75.9 2.0 <0.0001 0.5364 1041 (126) Feb. 04 40.7 3.4 <0.0001 0.3313 587(36) Mar. 04 14.5 2.8 0.0022 0.3991 373 (22) Apr. 04 17.6 4.8 0.0005 0.1794 288(39) location (Fig. 5). A seasonal migration is apparent in this path. After its release in July, the fish remained in the general vicinity of the release location for about 5 months, then abruptly in January and early February 2004, it migrated south approximately 200 km (straight line) from the release location to the southern most point of the path. In March and April, the fish nearly reversed direction and migrated to the north where it was recaptured. During the migration, the fish traveled an average of 6.4 km and maximum of 17.3 km per verti- cal excursion. The swimming speed which minimized the distance between the final migration path position and the reported capture location was 47 cm/s or 1.4 body lengths per second (BL/s) for the 34-cm fish. The bottom depths predicted at the locations along the migration path were very similar to the bottom depths (maximum depth within each 24-h period) mea- sured by the archival tag (Fig. 6) and had a mean ab- solute difference of only 2.5 m, thus corroborating the predicted migration path. The maximum depth during this migration occurred at the beginning of March when the fish abruptly changed its migration direction from the south to the northeast. Although a migration path for the southern fish could not be formulated, depth data from the archival tag (Fig. 2B), indicated that the fish must have remained along the Alaska Peninsula and did not migrate west toward the continental slope, or into the central Bering Sea shelf. Because predicted bottom depths gradually decreased from 90 m to 10 m over the first 10 months at liberty, the fish could not have migrated toward the slope. In addition, abrupt changes in bottom depth such as the 10-m to 40-m increase from 16 through 17 June 2004 (Fig. 2B), indicated that the fish remained in an area of relatively steep bathymetry — an area that does not exist on the central shelf. Accuracy of tidal current prediction Overall, current direction was more accurately pre- dicted at the northern mooring site, where direction errors were less than 40 degrees during periods of the Nichol and Somerton: Tidal stream transport of Lepidopsetta polyxystra in the eastern Bering Sea 227 most frequently observed current speeds (10-40 cm/s; Fig. 7A). By comparison, errors of over 40 degrees were common at the southern mooring site, particularly when current speeds were less than 20 cm/s (Fig. 7B). Model prediction of cur- rent direction improved with increasing observed current speed at both north- ern and southern mooring sites (Fig. 7). Model estimates of current speed at the northern mooring displayed progressive underestimation of the observed speed with increasing speed (Fig. 8A). When the observed current speeds were 33 cm/ s, for example, the model underestimated these speeds by an average of 10 cm/s. At the southern mooring, observed speeds were underestimated at slow speeds and overestimated at faster speeds (Fig. 8B). Discussion Archival tag data from two northern rock sole released and recaptured in the east- ern Bering Sea provide evidence that selective tidal stream transport can be used to aid horizontal migration. Verti- cal excursions, although infrequent, were correlated to both diel and tidal factors. Not only did both fish undergo verti- cal excursions during mostly nighttime hours, the northern fish did so during select periods of the tidal cycle when tidal currents were moving in a particu- lar direction. The significance of current direction as a determinate of vertical movement for the northern fish indicated that the fish did not randomly leave the bottom, but did so only when the current was moving in a specific direction. The successful application of tidal current information to predict a migration path for the northern fish validates that at least some northern rock sole use tidal currents for transport, and also may indicate that their vertical excursions are conducted primarily for the purpose of horizontal migration. Our attempt to predict a migration path for the southern fish was unsuccessful for several reasons. First, the identification of vertical excursions was less certain because of the variable bathymetric contours in the area. Second, model estimates of current direction were considerably less accurate when com- pared to the measured current direction at the location of the southern fish than for the location of the northern fish. Finally, the assumption for the model was that all horizontal movements occurred only during periods when the fish left the bottom; however, for the southern Northern fish B Southern fish Day (2003) Figure 3 Consecutive nightly vertical excursions (shown as peaks) recorded for two female northern rock sole ( Lepidopsetta polyxystra ) tagged with archival tags in the eastern Bering Sea. Examples for both northern (A) southern fish ( B ) are presented. Nighttime periods are shaded. Model estimates of tide height (heavy line) at a position midway between fish release and recovery positions are provided to highlight the tidal fluctuation recorded by the tag and show that the timing of vertical excursions is related to tidal factors in addition to diel factors. fish there was evidence of additional horizontal move- ment. Gradual decreases in fish depth sometimes oc- curred during periods in which tidal fluctuations could be recognized in the depth record, indicating that the fish migrated to some extent while it remained on or close to the seafloor. With no way to account for these movements in the model, we were unable to accurately calculate the projection of the path. The estimated average swim speed of 1.4 BL/s (47 cm/s) that minimized the distance between the final 228 Fishery Bulletin 107(2) Available Selected current current Sept 03 Available Selected current current Jan 04 Figure 4 Monthly plots from September 2003 through April 2004 of predicted near-bottom tidal current directions for periods when a female northern rock sole ( Lepidopsetta polyxystra ) (the northern fish) underwent nightly vertical excursions (selected current), compared to all the available nightly near-bottom tidal current directions (available current). Arrow directions indicate the direction of the current and arrow lengths indicate the total number of hours spent in each direction. The available current is scaled from 0 to 50 hours (see upper leftmost diagram) and the selected current is scaled from 0 to 10 hours migration path position and the recovery position is considerably higher than values of 0.6 BL/s reported for European plaice and Japanese flounder (Pai'alichthys olivaceus) during tidally assisted migration (Kawabe et ah, 2004; Metcalfe et ah, 1990). We offer several expla- nations. First, estimates of current speed in the model were lower than actual current speeds, a bias that would inflate our estimated swimming speed. Second, the northern fish may have been capable of maintaining a more consistent direction; that is, more aligned to the destination of the migration than the current directions. We assumed the fish traveled in the average direction of the current during each excursion event. However, if the fish had a predetermined destination and the ability to navigate, it may have actively deviated slightly from the average current direction. More efficient overall swim- ming directions would have reduced the overall dis- tance traveled, and again would have resulted in lower swimming speeds necessary to complete the migration path. Finally, some vertical excursions may not have been identified because of the frequency of collection of archival tag data (0.5 hour or 1 hour), or the fish may have migrated toward its destination without undergo- ing vertical excursions (e.g., it swam while near the bot- tom). Evidence of this type of movement was clear for the southern fish, which appeared to use tidal currents, but its horizontal migration was not limited to periods of vertical excursions. The northern fish may also have migrated without undergoing vertical excursions, but because the bathymetric terrain was fairly flat on the central eastern Bering Sea shelf, such movements could not be detected. Although we demonstrate the preference for night- time vertical activity for only two adult northern rock Nichol and Somerton: Tidal stream transport of Lepidopsetta polyxystra in the eastern Bering Sea 229 172°W 168°W 164°W 160°W Migration path of one female northern rock sole ( Lepidopsetta polyxystra ), tagged and recaptured in the eastern Bering Sea, released in 2003 and captured in 2004. The migration path is based on current velocity vectors predicted during periods when the northern fish underwent excursions away from the bottom. Circles identify the release and recovery positions. Gray lines with numbers indicate the bathymetric contours (m). Months are indicated by the alternating black and white paths. sole, this behavior is common among various flatfish species and juvenile northern rock sole. Nighttime ver- tical excursions have been reported for a variety of flatfishes from postlarval through adult stages (e.g., De Veen, 1967; Weinstein et ah, 1980; Cadrin and West- wood, 2004; Hunter et ah, 2004). Nighttime periods are thought to offer flatfishes a reduced risk of predation by visual predators (Burrows, 1994). In laboratory ex- periments, the swimming activity of juvenile northern rock sole (20-40 mm TL) away from the bottom oc- curred most often during nighttime (Hurst and Duffy, 2005). This activity involved vertical excursions to the surface, followed by horizontal swimming and gliding. It follows that, like adult northern rock sole, juveniles undergo vertical excursions away from the seafloor for the purpose of horizontal migration. Considering northern rock sole juveniles inhabit areas with tidal influence, it follows that they also use tidal currents for transport. Although it is unlikely that small juveniles migrate extensive distances, as some adults do, juve- niles may use tidal current for short-term migrations as a mechanism to locate better feeding grounds within nursery areas (Hurst and Duffy, 2005). We believe that at least some adult northern rock sole employ this 230 Fishery Bulletin 107(2) Month Figure 6 Bottom depths derived from an archival tag attached to a female northern rock sole (Lepidopsetta polyxystra) compared with bottom depths derived from the model- estimated fish migration path and bathymetric data. The fish was at liberty in the eastern Bering Sea from 2003 to 2004 (northern fish). feeding strategy, albeit on a larger spatial scale. The so called “feeding months” for northern rock sole in the eastern Bering Sea reportedly occur during summer when they disperse across the shelf after aggregating for spawning during winter and spring (Fadeev, 1965; Shubnikov and Lisovenko, 1964). This feeding period may well be represented by the first five months that the northern fish was at liberty ( July-November), dur- ing which migration was less frequent and not in a consistent direction. As suggested by the migration path presented here, at least some adult northern rock sole undergo vertical excursions for the purpose of tidally assisted horizontal migration, as opposed to vertical movements into the water column for feeding or spawning. Both juvenile and adult northern rock sole feed during daylight hours, but rarely during the night (Corcobado Onate, 1991; Hurst et al., 2007) when off-bottom swimming occurs. Unlike other eastern Bering Sea flatfish species such as arrowtooth flounder ( Atheresthes stomias), which feed high in the water column (Yang, 1995), northern rock sole feed close to the bottom. Northern rock sole feed almost exclusively on benthic invertebrates, such as polychaete worms and other marine worms (Corco- bado Onate, 1991; Lang et al., 1995; McConnaughey and Smith, 2000). In addition, northern rock sole likely do not leave the bottom to spawn because, along with their congener, southern rock sole (L. bilineata), they are the only northeast Pacific flatfishes to spawn de- mersal adhesive eggs (Matarese et al. 1989; Stark and Somerton, 2002). If most northern rock sole prefer to undergo vertical excursions during the night, winter offers greater op- portunity to travel in a preferred direction because of the increased hours of darkness. For the northern fish, vertical excursions were most frequent during January, when it travelled in a southerly direction. Southerly directed tidal currents were sometimes available at two different periods within a single night because of the semidiurnal nature of the tides (e.g., a full clockwise rotation of tidal currents every 12 hours). Hunter et al. (2004b) noted similar nocturnal behavior for Euro- pean plaice in the North Sea during winter when two ’’transporting tides” sometimes occurred within a night because of the longer periods of darkness lasting up to 15 hours. The northern rock sole vertical movements examined here appear shorter in both duration and extent (Table 1) in comparison with other flatfish for which vertical behavior has been studied, with the exception of yellow- tail flounder ( Limanda ferruginea ) whose excursions av- erage 1.5 hours and 6 m off bottom (Cadrin and Moser, 2006). Some European plaice ( P . platessa ) in the North Sea reportedly spend from 6 hours to 12 hours a night swimming in midwater during winter months (Hunter et al., 2004b). Common sole (Solea solea) in the North Sea are thought to use the upper half of the water col- umn for selective tidal stream transport during which Nichol and Somerton: Tidal stream transport of Lepidopsetta polyxystra in the eastern Bering Sea 231 140 120 - 100 - <10 10-20 20-30 30-40 40-50 >50 B 5. 140 § £ E 120 ■O CD > £ ioo <10 10-20 20-30 30-40 40-50 >50 Current speed (cm/s) Figure 7 Angle difference between observed and predicted (model) near- bottom tidal current directions at the mooring site near the northern fish (A) and at the mooring site near the southern fish (B) in the eastern Bering Sea, plotted at different observed current speeds. The y-axis is expressed as the mean absolute value of the angle difference between observed and predicted current directions (in degrees). Error bars indicate 25th and 75th percentiles. Percentages next to the means (circles) indicate the percentage of time that each grouping of observed current speed represented during the 193 days of collection. Figure 8 Comparison of observed and predicted (model) near-bottom tidal current speeds at the northern (A) and southern (B) mooring sites in the east- ern Bering Sea. Predicted speed is presented as the component of the model speed projected in the observed direction, MS0 cos(0), where MS0 = the model-predicted speed in the observed direction, and d = the angle difference between observed and predicted current directions. The line indicates a 1:1 ratio. Model estimates of speed do not include residual (e.g., baroclinic) effects that may have been present in observed current speeds. they frequently approach the surface (De Veen, 1967; Greer Walker et al., 1980). By comparison, the northern rock sole examined here only occasionally approached the surface; most vertical excursions occurred in the bottom half of the water column. Even during periods of active migration (e.g., January; northern fish), ver- tical excursions averaged only 2.6 hours in duration and were a maximum extent of 11.6 m away from the bottom. These vertical excursions could be of interest to fishery managers if they affect fish availability to bottom trawl surveys (Hunter et al., 2004b). However, the northern rock sole excursions were particularly infrequent during summer daylight hours when the bottom trawl surveys of the eastern Bering Sea are 232 Fishery Bulletin 107(2) conducted (Acuna and Lauth, 2008). From the data on the two fish examined here, northern rock sole remain on the bottom 99.8% of the time during summer (June and July) daylight hours. Although we could not predict a migration path for the southern fish, it was apparent that the migration pattern differed between the two fish. The northern fish clearly used tidal currents to facilitate a south- ern migration to deeper water during winter and a migration back north during spring. These movements are consistent with the seasonal spawning and post- spawning migrations suggested by Fadeev (1965) and Shubnikov and Lisovenko (1964). As with some Euro- pean plaice, which migrate south to warmer waters for spawning (Hunter et al., 2004a), northern rock sole that reside on the northern part of the eastern Ber- ing Sea shelf during summer (i.e., the northern fish) may require a migration to more southern or deeper waters to reach temperatures suitable for spawning. Adult rock sole (likely L. polyxystra ) from the western Bering Sea also undergo a migration to deeper water in winter, and do so presumably to avoid temperatures below 0°C (Shvetsov, 1979). Temperatures experienced by both fish decreased during winter months but sta- bilized to about 2°C in February and March. Had the northern fish stayed in the vicinity of its release, it would have experienced bottom temperatures below 0°C in February, as recorded by instruments at the northern oceanographic mooring site. The southern fish also underwent nighttime vertical excursions that were tidal in nature, but unlike the northern fish, there was no indication of a spawning migration; excursion frequency did not increase before the known spawning season (winter-spring), and depth records indicated no repeatable pattern from one winter (2004) to the next (2005). It is logical to assume that the extent of migrations is dependent on the proximity of feeding and spawning locations. Thus, the northern fish may require a directed seasonal migration to reach a vi- able spawning location, whereas the southern fish can remain resident if suitable feeding and spawning loca- tions are within close proximity. If the northern fish migrated south for the purpose of spawning, the southern extent of the migration route may have been a spawning location. We can infer from the spatial distribution of the fishery for roe of northern rock sole — a fishery that operates in the eastern Bering Sea during February and March just before the spawn- ing season (Stark and Somerton, 2002; Wilderbuer and Nichol, 2007) — that spawning aggregations occur over a wide area of the central and outer continental shelf extending from Unimak Island to west of the Pribilof Islands. This distribution overlaps with the southern point of the migration path. The eastern Bering Sea shelf offers a multitude of possibilities for tidally assisted transport, and the dis- tribution range that individuals seasonally inhabit may partly depend on the nature of the tidal currents. Based on tidal current ellipses for the M2 tidal constituent in the eastern Bering Sea, tidal currents are rotary in nature over the majority the shelf area south of lati- tude 60°N but become more bidirectional (i.e., 60- and 240-degrees) close to the Alaska Peninsula and into Bristol Bay (Pearson et al., 1981; Kowalik, 1999). Be- cause the northern fish inhabited the central part of the eastern Bering Sea continental shelf, opportunities for selective tidal stream transport were available in all directions, thus enabling a round-trip migration. By comparison, the southern fish resided along the Alaska Peninsula; therefore opportunities for selective tidal stream transport were limited to northeasterly and southwesterly directions. Fish that undergo migrations in northeasterly and southwesterly directions could use selective tidal stream transport over much of the eastern Bering Sea shelf. Adult yellowfin sole ( Limanda aspera), for example, are known to migrate annually in a northeasterly direction more than 500 km from winter grounds west and southeast of the Pribilof Islands to nearshore summer spawning grounds in Kuskokwim and Bristol bays ( Wakabayashi, 1989). Given the extent of this migration and the availability of tidal currents, it is reasonable to assume that yellowfin sole also use selective tidal stream transport. Results presented here provide the first known evi- dence of selective tidal stream transport among aquatic animals in the eastern Bering Sea. Among larval flat- fish in the eastern Bering Sea, including northern rock sole, passive forms of transport involving wind-driven surface currents and geostrophic flow have been shown to contribute to their horizontal distribution and like- lihood of survival (Wilderbuer et al., 2002; Lanksbury et al., 2007). The contribution of more active forms of transport such as selective tidal stream transport may become evident as more is learned about the verti- cal migration behavior of larvae. Evidence that larval northern rock sole as small as 8 mm can regulate their depth in the water column (Lanksbury et al., 2007) indicates that selective tidal stream transport is a pos- sibility. As we learn about how adult, juvenile, and larval fishes use tidal currents for migration, the need becomes evident for more accurate tide-prediction mod- els that can be used for modeling fish migration. Such models should become available in the near future with the completion of a baroclinic tide model of the eastern Bering Sea. Acknowledgments We thank D. Rachel and P. Stabeno for providing tidal information at Pacific Marine Environmental Labora- tory (PMEL) mooring sites, and for providing insight concerning tides in the eastern Bering Sea. A. Greig, S. Kotwicki, and J. Benson provided technical support on use of ArcView and circular trigonometry. T. Wilderbuer and J. Gauvin provided productive discussion concern- ing flatfish spawning and migration. Finally, K. Baily, T. Hurst, T. Wilderbuer, J. Lee, S. Cadrin, and three anonymous reviewers improved the manuscript with insightful suggestions. Nichol and Somerton: Tidal stream transport of Lepidopsetta polyxystra in the eastern Bering Sea 233 Literature cited Acuna, E., and R. R. Lauth. 2008. Results of the 2007 eastern Bering Sea conti- nental shelf bottom trawl survey of groundfish and invertebrate resources. NOAA Tech Memo. NMFS- AFSC-181, 195 p. Arnold, G. R, and B. H. Holford. 1995. A computer simulation model for predicting rates and scales of movement of demersal fish on the European continental shelf. ICES J. Mar. Sci. 52:981-990. Burrows, M. T. 1994. Foraging time strategy of small juvenile plaice: a laboratory study of diel and tidal behavior patterns with Artemia prey and shrimp predators. Mar. Ecol. Prog. Ser. 115:31-39. Cadrin, S. X., and J. Moser. 2006. Partitioning on-bottom and off-bottom behav- ior: a case study with yellowtail flounder off New England. ICES CM 2006/Q:14. Cadrin, S. X., and A. Westwood. 2004. The use of electronic tags to study fish move- ment: a case study with yellowtail flounder off New England. ICES CM 2004/K:81. Champalbert, G., C. Macquart-Moulin, G. Patriti, and L. Le Direach-Boursier. 1992. Light control of vertical movements of larvae and juvenile sole (Solea solea L.). Mar. Behav. Physiol. 19:263-283. Corcobado Onate, E. 1991. Food and daily ration of the rock sole Lepidop- setta bilineata (Pleuronectidae) in the eastern Bering Sea. Mar. Biol. 108:185-191. De Veen, J. F. 1967. On the phenomenon of Soles ( Solea solea L.) swim- ming at the surface. J. Cons. Cons. Int. Explor. Mer 31:207-236. Egbert, G. D., A. F. Bennett, and M. G. G. Foreman. 1994. TOPEX/Poseidon tides estimated using a global inverse model. J. Geophys. Res. 99:24,821-24,852. Egbert, G. D., and S. Y. Erofeeva. 2002. Efficient inverse modeling of barotropic ocean tides. J. Atmos. Ocean. Tech. 19:183-204. Fadeev, N. S. 1965. Comparative outline of the biology of flat- fishes in the southeastern part of the Bering Sea and condition of their resources. Translated by Isr. Prog. Sci. Trans., 1968. In Soviet fisheries investigations in the northeast Pacific, part 4 (P. A. Moiseev, ed.), p. 112-129. [Available from U.S. Dep. Commer., Natl. Tech. Inf. Serv., Springfield, VA, as TT 67- 51206.] Forward, R. B., and R. A. Tankersley. 2001. Selective tidal- stream transport of marine animals. Oceanogr. Mar. Biol. Annu. Rev. 39:305- 353. Fox, C. J., P. McCloghrie, E. F. Young, and D. M. Nash. 2006. The importance of individual behavior for success- ful settlement of juvenile plaice (Pleuronectes platessa L.): a modeling and field study in the eastern Irish Sea. Fish. Oceanogr. 15:301-313. Gibson, R. N. 2003. Go with the flow: tidal migration in marine animals. Hydrobiologia 503:153-161. Greer Walker, M., J. D. Riley, and L. Emerson. 1980. On the movement of sole ( Solea solea) and dogfish (Scyliorhinus canicula) tracked off the east Anglian coast. Neth. J. Sea Res. 14:66-77. Harden Jones, F. R., G. P. Arnold, M. Greer Walker, and P. Scholes. 1979. Selective tidal stream transport and migration of plaice (Pleuronectes platessa L.) in the southern North Sea. J. Cons. CIEM 38:331-337. Hunter, E., J. D. Metcalfe, G. P. Arnold, and J. D. Reynolds. 2004a. Impacts of migratory behavior on population structure in North Sea plaice. J. Aquat. Ecol. 73:377— 385. Hunter, E., J. D. Metcalfe, C. M. O’Brien, G. P. Arnold, and J. D. Reynolds. 2004b. Vertical activity patterns of free-swimming adult plaice in the southern North Sea. Mar. Ecol. Prog. Ser. 279:261-273. Hurst, T. P, and T. A. Duffy. 2005. Activity patterns in northern rock sole are medi- ated by temperature and feeding history. J. Exp. Mar. Biol. 325:201-213. Hurst, T. P., C. H. Ryer, J. M. Ramsey, and S. A. Haines. 2007. Divergent foraging strategies of three co-occurring north Pacific flatfishes. Mar. Biol. 151:1087-1098. Kawabe, R., Y. Naito, K. Sato, K. Miyashita, and N. Yamashita. 2004. Direct measurement of swimming speed, tailbeat, and body angle of Japanese flounder (Paralichthys olivaceus). ICES J. Mar. Sci. 61:1080-1087. Kowalik, Z. 1999. Bering Sea tides. In Dynamics of the Bering Sea (T. R. Loughlin and K. Ohtani, eds.), p. 93-127. Univ. Alaska Sea Grant Press., AK-SG-99-03, Fairbanks, AK. Kuipers, B. 1973. On the tidal migration of young plaice ( Pleuro- nectes platessa) in the Wadden Sea. Neth. J. Sea Res. 6:376-388. Lang, G. M., P A. Livingston, and B. S. Miller. 1995. Food habits of three congeneric flatfishes: Yellowfin sole (Pleuronectes asper), rock sole (P. bilineatus), and Alaska plaice (P. quadrituberculatus) in the eastern Bering Sea. In Proceedings of the international sym- posium on North Pacific flatfish, p. 225-245. Univ. Alaska Sea Grant Press, Fairbanks, AK. Lanksbury, J. A., J. T. Duffy-Anderson, K. L. Mier, M. S. Busby, and P. J. Stabeno. 2007. Distribution and transport patterns of northern rock sole, Lepidopsetta polyxystra, larvae in the south- eastern Bering Sea. Prog. Oceanogr. 72:39-62. Matarese, A. C., A. W. Kendall, D. M. Blood, and B. M. Vinter. 1989. Laboratory guide to the early life history stages of northeast Pacific fishes. NOAA Tech. Rep. NMFS 80, 652 p. McConnaughey, R. A., and K. R. Smith. 2000. Associations between flatfish abundance and sur- ficial sediments in the eastern Bering Sea. Can. J. Fish. Aquat. Sci. 57:2410-2419. Metcalfe, J. D., G. P. Arnold, and P. W. Webb. 1990. The energetics of selective tidal stream transport: an analysis for plaice tracked in the southern North Sea. J. Mar. Biol. Assoc., U K. 70:149-162. Metcalfe, J. D., E. Hunter, and A. A. Buckley. 2006. The migratory behavior of North Sea plaice: Cur- rents, clocks and clues. Mar. Freshw. Behav. Physiol. 39:25-36. Pearson, C. A., H. O. Mofjeld, and R. B. Tripp. 1981. Tides of the eastern Bering Sea shelf. In The eastern Bering Sea shelf: oceanography and resources 234 Fishery Bulletin 107(2) (D. W. Hood, and J. A. Calder, eds.), vol. 1, p. 111- 130. Office of Marine Pollution Assessment, NOAA, Univ. Washington Press, Seattle, WA. Rijnsdorp, A. D., and M. van Stralen. 1985. Selective tidal transport of North Sea plaice larvae Pleuj'onectes platessa in coastal nursery areas. Trans. Am. Fish. Soc. 114:461-470. Shubnikov, D. A., and L. A. Lisovenko. 1964. Data on the biology of rock sole of the southeast- ern Bering Sea. Translated by Isr. Prog. Sci. Trans., 1968. In Soviet fisheries investigations in the northeast Pacific, part 2 (P. A. Moiseev, ed.), p. 220-226. [Avail- able from U.S. Dep. Commer., Natl. Tech. Inf. Serv., Springfield, VA, as TT 67-51204.] Shvetsov, F. G. 1979. Distribution and migrations of the rock sole, Lepi- dopsetta bilineata bilineata, in the region of the Okhotsk Sea coast of Paramushir and Shumshu Islands. J. Icthyol./Vopr. Ikhtiol. 18:56-62. Stark, J. W., and D. A. Somerton. 2002. Maturation, spawning and growth of rock soles off Kodiak in the Gulf of Alaska. J. Fish. Biol. 61:417- 431. Venables W. N., and B. D. Ripley. 1994. Modern applied statistics with S-plus. Springer, New York, 462 p. Wakabayashi, K. 1989. Studies on the fishery biology of yellowfin sole in the eastern Bering Sea. [In Jpn., Engl, summ.] Bull. Far Seas Fish. Res. Lab. 26:21-152. Walsh, S. J., and M. J. Morgan. 2004. Observations of natural behavior of yellowtail flounder derived from data storage tags. ICES J. Mar. Sci. 61:1151-1156. Weinstein, M. P., S. L. Weiss, R. G. Hodson, and L. R. Gerry. 1980. Retention of three taxa of postlarval fishes in an intensively flushed tidal estuary, Cape Fear River, North Carolina. Fish. Bull. 78:419-435. Wilderbuer, T. K., A. B. Hollowed, W. J. Ingraham Jr., P. D. Spencer, M. E. Conners, N. A. Bond, and G. E. Walters. 2002. Flatfish recruitment response to decadal climatic variability and ocean conditions in the eastern Bering Sea. Prog. Oceangr. 55:235-247. Wilderbuer, T. K., and D. G. Nichol. 2007. Northern rock sole. In Stock assessment and fishery evaluation report for the groundfish resources of the Bering Sea/Aleutian Islands region, chapter 7, p. 627-686. [Available from North Pacific Fishery Man- agement Council, 605 West 4th Ave, Suite 306, Anchor- age, AK 99501.] Yang, M. 1995. Food habits and diet overlap of arrowtooth flounder ( Atheresthes stomias ) and Pacific halibut ( Hippoglossus stenolepis) in the Gulf of Alaska. In Proceedings of the international flatfish symposium; October 1994, Anchorage, Alaska, p. 205-223. 235 Prosomal-width-to-weight relationships in American horseshoe crabs ( Limulus polyphemus ): examining conversion factors used to estimate landings Brian R. Murphy1 Email address for contact author: L|g85@cornell.edu 1 Department of Fisheries and Wildlife Sciences Virginia Polytechnic Institute and State University 100 Cheatham Hall Blacksburg, Virginia 24061 Present address for contact author: New York Sea Grant Stony Brook University 146 Suffolk Hall Stony Brook, New York 11794-5002 2 Department of Natural Sciences Fordham University, College at Lincoln Center 1 1 3 West 60,h Street New York, New York 10023 3 Department of Ecology, Evolution, and Natural Resources Cook College, Rutgers University New Brunswick, New Jersey 08901 Abstract — Horseshoe crabs ( Limu- lus polyphemus) are valued by many stakeholders, including the commer- cial fishing industry, biomedical com- panies, and environmental interest groups. We designed a study to test the accuracy of the conversion fac- tors that were used by NOAA Fish- eries and state agencies to estimate horseshoe crab landings before man- datory reporting that began in 1998. Our results indicate that the NOAA Fisheries conversion factor consis- tently overestimates the weight of male horseshoe crabs, particularly those from New England populations. Because of the inaccuracy of this and other conversion factors, states are now mandated to report the number (not biomass) and sex of landed horse- shoe crabs. However, accurate esti- mates of biomass are still necessary for use in prediction models that are being developed to better manage the horseshoe crab fishery. We recommend that managers use the conversion fac- tors presented in this study to convert current landing data from numbers to biomass of harvested horseshoe crabs for future assessments. Manuscript submitted 11 June 2008. Manuscript accepted 14 January 2009. Fish. Bull. 107:235-243 (2009). The views and opinions expressed or implied in this article are those of the author and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Larissa J. Graham (contact author)1 Mark L. Botton2 David Hata Robert E. Loveland3 Horseshoe crabs ( Limulus polyphe- mus) are considered a multiple-use resource. They are valued by many stakeholders, including the commer- cial fishing industry, biomedical com- panies, and environmental interest groups (Berkson and Shuster, 1999). Horseshoe crabs are commercially harvested and sold as bait for whelk ( Busycon spp. and Busycotypus spp.) and American eel (Anguilla rostrata) fisheries. This species is also gath- ered for biomedical companies because its copper-containing blood is used to create a pharmaceutical product, Lim- ulus amoebocyte lysate (LAL) that is used to detect pathogenic endotoxins on medical devices and in injectable drugs (Novitsky, 1984; Mikkelsen, 1988; Levin et ah, 2003). The mor- tality associated with the handling and bleeding of horseshoe crabs is minimized (i.e., 8-20% [Rudloe, 1983; Kurz and James-Pirri, 2002; Walls and Berkson, 2003; Hurton and Berk- son, 2004]) because the animals are required to be returned to the water within 72 hours. Horseshoe crabs are ecologically important because their eggs serve as a food source for migrat- ing shorebirds most notably in Dela- ware Bay (Tsipoura and Burger, 1999; Botton et al., 2003; Karpanty et al., 2006; Haramis et al., 2007). In 1998, a fishery management plan was developed for the horseshoe crab. However, before this plan, most states did not require the manda- tory reporting of harvested horse- shoe crabs. NOAA Fisheries collected commercial landing data by state, year, and gear type, but these data were incomplete and disjunct. To es- timate reference period (or a basis for reductions in landing data), the Horseshoe Crab Technical Commit- tee asked state agencies to provide their best estimate of the number of horseshoe crabs harvested before 1998. These numbers were converted to pounds using various conversion factors. The number of horseshoe 236 Fishery Bulletin 107(2) crabs harvested in Delaware and Virginia waters were converted to biomass using conversion factors derived from fishery-independent and fishery-dependent data (i.e., Delaware: 1.05 kg/male, 2.32 kg/female, 1.69 kg/ combined catch; Virginia: 1.8 kg/horseshoe crab or 2.27 kg/horseshoe crab depending on the composition of the catch). The landing data from all other states were converted to pounds using a NOAA Fisheries conversion factor (i.e., 1.21 kg/horseshoe crab). These data have since been used to generate estimates of total landings, to set state-by-state quotas, and to manage the stock (Fig. 1). Once the horseshoe crab fishery management plan was initiated, all landings were required to be re- ported by sex, harvest method, and the number and pounds of harvested horseshoe crabs. However, many fishermen reported their catch in numbers of har- vested horseshoe crabs, and state agencies used con- version factors to convert harvests from numbers to pounds. Because of the uncertainty in these conver- sion factors and resulting biomass estimates, state agencies are no longer required to report the pounds of horseshoe crabs landed. The Atlantic States Marine Fisheries Commission (ASMFC) and state agencies now assess and manage stocks using only the number of horseshoe crabs (not pounds) harvested. ASMFC currently uses trend analysis to manage horseshoe crab populations, but numerous prediction models are being developed for future, more accurate management. For some of these models, landing data are required to be reported in pounds, not numbers. Because all state landings are currently reported by numbers of landed horseshoe crabs, conversion fac- tors need to be derived to estimate pounds of landed horseshoe crabs. The availability of accurate conversion factors will serve as a factor in choosing an appropriate model to better manage horseshoe crab populations. The objective of our study was to derive prosomal- width-to-weight equations to calculate alternative sex- specific conversion factors based on the average width of horseshoe crabs from each state. We also tested the NOAA Fisheries conversion factor by comparing the observed total biomass of horseshoe crabs to the total biomass that was estimated with the conversion factor. Materials and methods Data collection Data were collected during three spawning surveys in the Mid-Atlantic (i.e., Delaware Bay, NJ, sampled in 1997 and 2000 [n = 379]; Raritan Bay, NJ, sampled in 1988 [/r = 297] ) and southern New England (i.e., Great Bay, NH, sampled in 1988 [/r = 131]) and from the Del- aware commercial fishery (i.e., Delaware Bay, DE, sampled in 1999, 2003, and 2004 [n = 348]) (Fig. 2). The sex, prosomal width (PW; to the nearest 1 mm), and weight (to the nearest 10 g) were recorded from a sample of individuals that were collected from the vicin- ity of each breeding beach. During spawning surveys, animals were collected as either mated pairs (a male coupled to a female) or as unpaired (or satellite males) because previous studies (Botton and Loveland, 1989) have shown that there is no significant size difference between unattached males within a population. The majority of the samples collected from the commercial fishery were harvested by hand during spawning events. All samples were mature individu- als because only mature horse- shoe crabs visit beaches during spawning events. Measurements also were re- corded for horseshoe crabs in coastal waters (i.e., within 12 nautical miles from shore) be- tween New York and Virginia (Fig. 2). In September, Octo- ber, and November of 2005 and 2006, 743 individuals were col- lected and measured during the Horseshoe Crab Research Center (HCRC) trawl survey (for meth- ods see Hata and Berkson, 2004). In June of 2006, an additional 182 horseshoe crabs were sam- pled aboard a commercial trawl vessel harvesting crabs for a bio- medical company off the coast of Ocean City, Maryland. Trawling gear, specifically a flounder net, was used to collect all horseshoe Graham et at: Prosomal-width-to-weight relationships for Limulus polyphemus 237 72°0'0"W 71°0'0"W 70°0'0"W Figure 2 Sites sampled for horseshoe crabs during the Horseshoe Crab Research Center (HCRC) trawl survey of inshore continental shelf waters between New York and Virginia (n = 50 sites) and from spawning surveys in New Jersey and Delaware (i.e. , Delaware Bay), and New Hampshire (i.e.. Great Bay), and from the Delaware commercial fishery (i.e., Delaware Bay). crabs. The ground gear on the flounder net was modi- fied with a Texas sweep, which has a chain line in- stead of a footrope, to effectively sample horseshoe crabs (Hata and Berkson, 2003; Hata and Berkson, 2004). We recorded prosomal width (to the nearest 1 mm), weight (to the nearest 10 g), sex, and maturity stage for all or a subsample of horseshoe crabs at each site. Maturity stage was classified into two groups: immature and mature. Male horseshoe crabs without modified pedipalps (claspers) were considered imma- ture and those with modified pedipalps were consid- ered mature (Hata and Berkson, 2004). Females with mating scars (i.e., indentations and abrasions on the dorsal surface of the opisthosoma resulting from the attached male) were categorized as mature. Maturity stage in newly molted females is not morphologically distinct, therefore some individuals had to be probed with an awl for evidence of eggs and determine the stage of maturity (Leschen et al., 2006). Females with eggs were categorized as mature (Hata and Berkson, 2004). Prosomal-width-to-weight relationship We log-transformed the PW and weight measurements collected during the HCRC trawl survey and used a general linear model (PROC GLM, SAS, vers. 9.1, SAS Inst., Inc., Cary, NC) to test for significant differences in the PW, weight, and PW-weight relationship between sexes and maturity stages. The P-value of each family of comparisons was adjusted using a Bonferroni correction to protect the experimental-wise error rate. We combined all data (i.e., three spawning surveys, Delaware commercial fishery, HCRC trawl survey) to develop PW-weight regression equations for each group (i.e., mature males, mature females, immature males, 238 Fishery Bulletin 107(2) and immature females) of horseshoe crabs using the form loge(VFf) = loge(fW)-a + loge(6), (1) where Wt = weight of a horseshoe crab (kg); PW = prosomal width (mm); a = slope; and b = y-intercept (PROC REG, SAS). We could not develop equations for each group of horseshoe crabs by state because of the small sample size collected from some states. Testing current and developing alternative conversion factors The predictive accuracy of the NOAA Fisheries conver- sion factor was tested using data collected from four data sets: the three spawning surveys and the Delaware commercial fishery. We calculated total biomass for each sample using the NOAA Fisheries conversion factor and then compared it to the total observed biomass for each sample. We used various data sets (i.e., three spawning sur- veys, the Delaware commercial fishery, the HCRC trawl survey, unpublished data) and previously published studies to generate the average PW and weight for male and female horseshoe crabs from each state (Table 1). For some states, the average weight of horseshoe crabs was not available, and therefore we used the PW-weight equations that were derived from this study to estimate the average weight of horseshoe crabs based on an aver- age measured prosomal width. For states where average weight data were available, we compared the observed weight to the estimated weight (i.e., using PW-weight equation) to determine the accuracy of the PW-weight equations. Results Prosomal-width-to-weight relationship The average weight differs between male and female horseshoe crabs. Mature female horseshoe crabs were significantly larger (i.e., prosomal width; df=l, 346; F= 1488.03; P< 0.0001) and heavier (df=l, 346; F=2245.72; PcO.0001) than mature male horseshoe crabs. The weight of horseshoe crabs was significantly different among sex and maturity stages (df=7, 924; F=6.86; P=0.0090; Table 2). Significant differences did not occur in the PW-weight relationship of mature male and mature female horseshoe crabs (df=3, 577; F=2.19; P=0.1396; Table 2); however, when comparing only horseshoe crabs of overlapping size ranges (PW=181-292 mm; weight= 0.88-3.14 kg), the PW-weight relationship of mature female horseshoe crabs was significantly dif- ferent than that of mature males (df=3, 626; F=8.21; P=0.0043). Separate PW-weight equations were developed for all females, mature females, immature females, all males, mature males, and immature males (Table 3). The derived PW-weight equations were used to esti- mate an average weight of horseshoe crabs from each state based on the observed prosomal width (Table 1). However, we used only the PW-weight equation derived for mature horseshoe crabs (i.e., one for mature males and one for mature females) to estimate weight because the PW-weight relationship was significantly different between sexes of mature horseshoe crabs, and the com- mercial fishery is directed only at mature horseshoe crabs. The estimated average weight of both male and female horseshoe crabs with the derived PW-weight equations was relatively accurate compared to the ob- served average weight for each state (Table 1). Testing current and developing alternative conversion factors The conversion factor used by NOAA Fisheries (i.e., 1.21 kg/horseshoe crab) consistently overestimated the total weight of horseshoe crabs collected during spawn- ing surveys and from the Delaware commercial fishery (Table 4). For female horseshoe crabs from Mid-Atlantic populations (i.e., Delaware Bay and Raritan Bay), this conversion factor provided a relatively close estimate of total weight. However, when estimating the total weight of male horseshoe crabs, the NOAA Fisheries conversion factor overestimated total weight. The weight of horse- shoe crabs from the New England population (i.e., Great Bay) was overestimated to the greatest degree, by more than 70% for both males and females. The average weight of a horseshoe crab also varies by location. Horseshoe crabs between Rhode Island and South Carolina are larger and heavier than horseshoe crabs from Maine, New Hampshire, Massachusetts, and Florida; and the conversion factors that have been used by most states reflect the differences in size and weight among states (Table 1). For those states where a single conversion factor has been used in the past to estimate the weight for both male and female horseshoe crabs (i.e., Maine, Rhode Island, Virginia, North Carolina, South Carolina, and Florida), the weight of at least one sex, in most cases the weight of female horseshoe crabs (Table 4) has been predicted inaccurately. Most states in the Mid-Atlantic have derived two conversion factors (i.e., one for each sex) that are relatively close to the average weight of mature horseshoe crabs collected within that area (Table 1). Discussion Female horseshoe crabs are much larger than male horseshoe crabs; therefore separate conversion factors should be used for each sex. Our results indicate that horseshoe crabs exhibit considerable sexual size dimor- phism with mature female horseshoe crabs being sig- nificantly larger and heavier than males. Males in any Graham et al.: Prosomal-width-to-weight relationships for Limulus polyphemus 239 240 Fishery Bulletin 107(2) Graham et al.: Prosomal-width-to-weight relationships for L/mulus polyphemus 241 Table 2 General linear model F-values and P-values for horseshoe crabs ( Limulus polyphemus) that were collected during the Horseshoe Crab Research Center trawl survey which sampled inshore continental shelf waters between New York and Virginia. Values are listed for all horseshoe crabs combined, immature and mature females, immature and mature males, mature females and males, and immature females and males. The prosomal-widthl PW )-to-weight relationship was analyzed for various combinations of sex and maturity stage (Mat). Significant interactions, after Bonferroni adjustment, are indicated by an asterisk. Variable All data (df=7, 924; n=925) Mature females vs. Immature females (df=3, 481;ti = 482) Mature males Immature females vs. vs. Immature males Immature males (df=3, 442; w =443 )(df= 3, 346; tj = 347) Mature females vs. Mature males (df=3, 577; 72 = 578) F P F P F P F P F P PW 3770.45 <0.0001* 2490.51 < 0.0001* 1456.77 <0.0001* 2849.41 <0.0001* 2056.93 <0.0001* Sex 0.00 0.9866 — — — — 6.17 0.0135* 3.46 0.0635 Mat 0.20 0.6582 1.46 0.2271 5.61 0.0183 — — — — PW x Sex 1.16 0.2815 - — — — 6.76 0.0097* 2.19 0.1396 PWxMat 1.59 0.2075 0.98 0.3234 5.51 0.0194 — — — — Sex x Mat 6.86 0.0090* — — — — — — — — PW* Sex* Mat 6.18 0.0131 — — — — — — — — Table 3 The number of individuals sampled (n), coefficient values ( a, b ), standard errors for coefficients (SE [a], SE [6] ), and correlation coef- ficient (r2) of the relationship between prosomal width and weight for horseshoe crabs (Limulus polyphemus ), loge( Wif)=log(,(PVTx o +loge(fi). Samples were collected during the Horseshoe Crab Research Center trawl survey (i.e., inshore continental shelf waters between New York and Virginia), spawning surveys (i.e.. New Jersey, Delaware, New Hampshire), and the commercial fishery (i.e., Delaware). All regressions are significant. n a b SE(a) SE(b) r2 Female (all) 1025 2.98 -15.71 0.02 0.10 0.96 Females (mature) 802 2.65 -13.85 0.04 0.21 0.86 Females (immature) 223 2.85 -15.10 0.05 0.23 0.95 Males (all) 1055 2.89 -15.39 0.02 0.12 0.94 Males (mature) 931 2.97 -15.80 0.02 0.13 0.94 Males (immature) 124 2.58 -13.81 0.10 0.50 0.85 population average about 80% of the prosomal width of the females (Shuster, 1979) and mature females are significantly heavier than mature males because of their larger size and added weight associated with numer- ous eggs within their prosomas (Leschen et al., 2006). Therefore, it is inappropriate to use the same conversion factor for both sexes. Conversion factors should also vary by state, to take into account the larger size and greater weight of horse- shoe crabs in Mid-Atlantic states. Horseshoe crabs from the middle Atlantic region are significantly larger than animals from Cape Cod Bay to Maine and those from the Gulf of Mexico (Shuster, 1979). Morphometries (Shuster, 1979; Riska, 1981), survey data on the dis- tribution of horseshoe crabs along the continental shelf (Botton and Ropes, 1987), and population genetic stud- ies (King et al., 2005) strongly indicate that there are geographically distinct breeding populations throughout the range. Some intermingling of populations occurs along the middle Atlantic coast, especially from New Jersey to Virginia (Swan, 2005), where much of the trawl-based fishery has been located. Because of this geographic variation, it is inappropriate to use the same conversion factor for horseshoe crabs from all states. The conversion factor that was used by NOAA Fisher- ies (i.e., 1.21 kg per horseshoe crab) to estimate refer- ence period landing data does not accurately estimate total biomass. From our results, it seems that reference period landing data were overestimated, especially in cases where the fishery could have been male-biased. The effects of this inaccurate conversion factor could have been further magnified in areas where the aver- age size and weight of horseshoe crabs is much smaller than that for Mid-Atlantic states, notably embayments from the northern (Cape Cod to Maine) and southern (Gulf of Mexico) parts of the distribtuion range of this 242 Fishery Bulletin 107(2) Table 4 The aggregate observed weight and estimated weight using the NOAA Fisheries conversion factor (i.e., 1.21 kg per animal) of horseshoe crabs ( Limulus polyphemus) collected during spawning surveys. The percent that the NOAA Fisheries conversion factor overestimates weight is also listed. Location Sex Aggregate observed weight (kg) Aggregate estimated weight (kg) Percent weight overestimated New Jersey (Delaware Bay) Female (rz =168) 446 448 0.3 Male (n=211) 237 563 58 Total (n=379) 683 1011 32 New Jersey (Raritan Bay) Female (n = 102) 231 272 15 Male (n = 195) 192 521 63 Total (n =297 ) 424 793 47 Delaware (Delaware Bay)i Female (n=261) 631 697 9 Male (n = 87) 90 232 61 Total (rc=348) 721 929 22 New Hampshire (Great Bay) Female (n=12) 7 31 77 Male (« = 119) 28 309 91 Total (n = 131) 35 341 90 1 S. Michels. Unpubl. data. 1999, 2003, 2004. Delaware Fish and Wildlife, P.O. Box 330, Little Creek, DE 19961. species. According to our analyses, a New England har- vest, composed of mostly male horseshoe crabs, would be the worst-case scenario for overestimating landings data when measured in pounds. To more accurately estimate reference period land- ings, biomass should be recalculated using state-specific conversion factors for each sex. However, determining the male-to-female ratios from landing data may be a challenge. Before 1998, participants in the fishery were not required to record the ratio of males to females among landed horseshoe crabs. It has been suggested that eel bait fishermen prefer to harvest females, be- cause of a chemical attractant associated with the eggs (Ferrari and Targett, 2003). In contrast, both male and female horseshoe crabs were used, as available, for the whelk fishery. Unfortunately, no data are available on the percentage of horseshoe crabs landed as bait for eels versus whelks, from which one might be able to deduce the sex ratio in the early commercial catches. Future estimates of the biomass of harvested horse- shoe crabs should incorporate the sex and location of horseshoe crab harvests. Use of geographically-appro- priate conversion factors for each sex would provide an accurate estimate of biomass despite the differing regulations among states. Some states have already derived their own sex-specific conversion factors, and most seem to provide an accurate representation of the average weight for male and female horseshoe crabs. States that have used one conversion factor to estimate the weight of both female and male horseshoe crabs (i.e., Maine, Rhode Island, Virginia, South Carolina, and Florida) either underestimate the weight of female horseshoe crabs or overestimate the weight of male horseshoe crabs. Although state agencies are no longer required to report landings in number and pounds, the conversion factors that have already been derived by state agencies may serve as a useful tool for accurately converting data to be used in prediction models. For states that have not developed accurate conversion fac- tors, the PW-weight equations derived from this study can be used to develop conversion factors based on the average width of male and female horseshoe crabs from that area. Besides providing a more accurate estimate of biomass, use of state-specific and sex-specific conver- sion factors is feasible for management purposes be- cause states are already required to report the location, sex, and number of horseshoe crabs harvested. At present, only very limited size and weight data are available for horseshoe crabs from North Carolina through northern Florida. Our PW-weight relationships for both sexes are very robust across a wide range of sizes, but could be further improved by the inclusion of horseshoe crab populations from this part of their range. Conclusion It is important to provide accurate biomass estimates of harvest data for future management purposes and, therefore, accurate conversion factors should be devel- oped. From the results of this study, it seems that the most practical approach to estimating landing data is to use state-specific conversion factors, one for females and one for males, based on the average weight of horse- shoe crabs from that area. Researchers should continue Graham et al : Prosomal-width-to-weight relationships for Limulus polyphemus 243 to collect data on the average PW of female and male horseshoe crabs to fine tune these conversion factors. The PW-equations derived from this study can be used to estimate weight based on an average prosomal width. In this way, the accuracy of these conversion factors could be improved, thereby providing better data for future management assessments. Acknowledgments We sincerely thank S. Michels, M. Beekey, and J. Mattei for generously providing unpublished data on horse- shoe crab populations from Delaware and Connecticut (listed in Table 1), and B. Spear for his contributions to this manuscript. J. Brust, R. Burgess, L. DeLancey, S. Doctor, S. Gerhart, L. Gillingham, P. Himchak, A. Leschen, C. McBane, T. Moore, M. Oates, S. Olszewski, and P. Thayer provided valuable information about state conversion factors. We also thank M. Davis, M. Duncan, R. Leaf, B. Ray, and A. Villamagna for their assistance with data analyses, manuscript reviews, and discussion of research ideas. Literature cited Berkson, J. and C. N. Shuster. 1999. The horseshoe crab: the battle for a true multiple- use resource. Fisheries 24:6-10. Botton, M. L., B. A. Harrington, N. Tsipoura, and D. Mizrahi. 2003. Synchronies in migration: Shorebirds, horseshoe crabs, and Delaware Bay. In The American horseshoe crab (C. N. Shuster Jr., R. B. Barlow, and H. J. Brock- mann, eds.), p. 5-32. Harvard Press, Cambridge, MA. Botton, M. L., and R. E. Loveland. 1989. Reproductive risk: high mortality associated with spawning by horseshoe crabs ( Limulus polyphemus) in Delaware Bay, USA. Mar. Biol. 101:143-151. Botton, M. L., and J. W. Ropes. 1987. Populations of horseshoe crabs, Limulus polyphemus , on the northwestern Atlantic continental shelf. Fish. Bull. 85:805-812. Ferrari, K. M., and N. M. Targett. 2003. Chemical attractants in horseshoe crab, Limulus polyphemus , eggs: The potential for an artificial bait. J. Chem. Ecol. 29:477-496. Haramis, G. M., W. A. Link, P. C. Osenton, D. B. Carter, R. G. Weber, N. A. Clark, M. A. Teece, and D. S. Mizrahi. 2007. Stable isotope and pen feeding trial studies confirm value of horseshoe crab eggs to spring migrant shorebirds in Delaware Bay. J. Avian Biol. 37:367-376. Hata, D., and J. Berkson. 2003. Abundance of horseshoe crabs ( Limulus polyphemus ) in the Delaware Bay area. Fish. Bull. 101:933-938. 2004. Factors affecting horseshoe crab Limulus poly- phemus trawl survey design. Trans. Am. Fish. Soc. 133:292-299. Hurton, L., and J. Berkson. 2004. Potential causes of mortality for horseshoe crabs (Limulus polyphemus) during the biomedical bleeding process. Fish. Bull. 104:293-298. James-Pirri, M. J., K. Tuxbury, S. Marino, and S. Koch. 2005. Spawning densities, egg densities, size structure, and movement patterns of spawning horseshoe crabs, Limulus polyphemus, within four coastal embayments on Cape Cod, Massachusetts. Estuaries 20:296- 313. Karpanty, S. M., J. D. Fraser, J. Berkson, L. J. Niles, A. Dey, and E. P. Smith. 2006. Horseshoe crab eggs determine red knot distribution in Delaware Bay. J. Wildl. Manag. 70:1704-1710. King, T. L., M. S. Eackles, A. P. Spidle, and H. J. Brockmann. 2005. Regional differentiation and sex-based disper- sal among populations of horseshoe crabs Limulus polyphemus. Trans. Am. Fish. Soc. 134:441-465. Kurz, W., and M. J. James-Pirri. 2002. The impact of biomedical bleeding on horseshoe crab, Limulus polyphemus, movement patterns on Cape Cod, Massachusetts. Mar. Freshw. Behav. Physiol. 35:261-268. Leschen, A. S., S. P. Grady, and I. Valiela. 2006. Fecundity and spawning of the Atlantic horse- shoe crab, Limulus polyphemus, in Pleasant Bay, Cape Cod. Mar. Ecol. 27:54-65. Levin, J., H. D. Hochstein, and T. J. Novitsky. 2003. Clotting cells and Limulus amoebocyte lysate: an amazing analytical tool. In The American horseshoe crab (C. N. Shuster, R. B. Barlow, H. J. Brockmann, eds.), p. 310-340. Harvard Univ. Press, Cambridge, MA. Mikkelsen, T. 1988. The secret in the blue blood. Science Press, Bei- jing, China. Novitsky, T. J. 1984. Discovery to commercialization: the blood of the horseshoe crab. Oceanus 27:13-18. Riska, B. 1981. Morphological variation in the horseshoe crab Limulus polyphemus. Evolution 35:647-658. Rudloe, A. 1983. The effect of heavy bleeding on mortality of the horseshoe crab, Limulus polyphemus, in the natural environment. J. Invertebr. Pathol. 42:167-176. Shuster, C. N. Jr. 1979. Distribution of the American horseshoe “crab,” Limulus polyphemus ( L . ) . In Biomedical applications of the horseshoe crab (Limulidae) (E. Cohen, ed.), p. 3-26. Alan R. Liss Inc., New York. Swan, B. L. 2005. Migrations of adult horseshoe crabs, Limulus poly- phemus, in the Middle Atlantic Bight: a 17-year tagging study. Estuaries and Coasts 28:28-40. Tsipoura N., and J. Burger. 1999. Shorebird diet during spring migration stopover on Delaware Bay. Condor 101:635-644. Walls, E. A., and J. Berkson. 2003. Effects of blood extraction on horseshoe crabs ( Limulus polyphemus). Fish. Bull. 101:457—459. 244 Abstract — The Pacific Rim population structure of chum salmon ( Oncorhyn - chus keta) was examined with a survey of microsatellite variation to describe the distribution of genetic variation and to evaluate whether chum salmon may have originated from two or more glacial refuges fol- lowing dispersal to newly available habitat after glacial retreat. Variation at 14 microsatellite loci was surveyed for over 53,000 chum salmon sampled from over 380 localities ranging from Korea through Washington State. An index of genetic differentiation, FST, over all populations and loci was 0.033, with individual locus values ranging from 0.009 to 0.104. The most genetically diverse chum salmon were observed from Asia, particularly Japan, whereas chum salmon from the Skeena River and Queen Char- lotte Islands in northern British Columbia and those from Washington State displayed the fewest number of alleles compared with chum salmon in other regions. Differentiation in chum salmon allele frequencies among regions and populations within regions was approximately 18 times greater than that of annual varia- tion within populations. A regional structuring of populations was the general pattern observed, with chum salmon spawning in different tribu- taries within a major river drainage or spawning in smaller rivers in a geo- graphic area generally more similar to each other than to populations in different major river drainages or geo- graphic areas. Population structure of chum salmon on a Pacific Rim basis supports the concept of a minimum of two refuges, northern and south- ern, during the last glaciation, but four possible refuges fit better the observed distribution of genetic varia- tion. The distribution of microsatellite variation of chum salmon on a Pacific Rim basis likely reflects the origins of salmon radiating from refuges after the last glaciation period. Manuscript submitted 28 August 2008. Manuscript accepted 22 January 2009. Fish. Bull. 107:244-260 (2009). The views and opinions expressed or implied in this article are those of the author and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Population structure of chum salmon ( Oncorhynchus keta) across the Pacific Rim, determined from microsatellite analysis Terry D. Beacham (contact author) John R. Candy Khai D. Le Michael Wetklo Email address for contact author: Terry.Beacham@dfo-mpo.gc.ca Fisheries and Oceans Canada, Pacific Biological Station, 3190 Hammond Bay Road Nanaimo, B. C., Canada V9T 6N7 Delineation of phylogenetically and adaptively distinct groups in the dis- tribution of chum salmon around the Pacific Rim may lead to conserva- tion of genetic diversity through an understanding of the origin and the evolutionary processes promoting and maintaining genetic differentiation. An evaluation of genetic variation in describing the population structure of salmonids, is a key component in the elucidation of management units or conservation units in a species and can be applied to manage fisheries exploiting specific stocks of salmon. Several methods of surveying genetic variation have been used to investi- gate regional and Pacific Rim varia- tion in chum salmon ( Oncorhynchus keta Walbaum). Allozymes have been used for a number of years to describe chum salmon population differentia- tion and structure (Okazaki, 1982a; Kijima and Fujio, 1982; Wilmot et ah, 1994; Efremov, 2001; Salmenkova et al., 2007). Variation in mitochondrial (mt) DNA has also been investigated (Park et ah, 1993; Sato et al., 2001, 2004), as has minisatellite variation (Taylor et al., 1994; Beacham, 1996). Non-mtDNA single nucleotide polymor- phisms have been examined (Smith and Seeb, 2008), as have microsat- ellites (Chen et al. 2005; Beacham et al. 2008a, 2008b, 2008c, 2009). Microsatellites are useful for evalu- ating fine-scale population structure in salmonids (Banks et al., 2000), and for investigating population structure around the Pacific Rim (Beacham et ah, 2006a, 2006b). Chum salmon display one of the widest spawning distributions of Pa- cific salmon. In Asia, chum salmon are distributed from Korea and Ja- pan in the south to the Arctic Ocean coast of Russia in the north; in North America, the distribution has histori- cally ranged from California in the south to the Beaufort Sea coast in the north, and as far east as the Mack- enzie River in the Arctic (Salo, 1991). After fry emerge from the gravel nest in the spring or are released from hatcheries, they generally move di- rectly to marine residence, first to estuaries, and later in the year to nearshore and offshore waters. Most individuals reside three to five years in the marine environment and then undertake spawning migrations gen- erally to their natal river beds. Chum salmon were likely fairly widely distributed along the Pacific Rim before the last major glaciation (McPhail and Lindsey, 1970). The advent of glaciation restricted the distribution of chum salmon to some major and perhaps minor refuges. Existing chum salmon population structure has been associated with colonization events following the last glaciation (Seeb and Crane, 1999). Modern populations were thought to have originated largely from a Ber- ing Sea refuge in the north and a Columbia River refuge in the south (McPhail and Lindsey, 1970). In Asia, Beacham et al. : Population structure of Oncorhynchus keta across the Pacific Rim 245 local refuges may also have been present in the Kam- chatka region (Varnavskaya et al., 1994), and in British Columbia, on the Queen Charlotte Islands and perhaps on coastal islands in the central coast region (Warner et al., 1982; Wood, 1995). Seeb and Crane (1999) indi- cated that existing populations from the Alaska Penin- sula south to Washington may have derived primarily from the southern refuge, whereas Asian and western Alaskan populations may have derived from a northern refuge. Microsatellite variation can be used to examine relationships between existing Pacific Rim population structure and proposed patterns of dispersal from gla- cial refuges. In the current study, we evaluated chum salmon dis- persal pathways from glacial refugia after glacial re- treat. In addition, we examined regional differentiation in allelic frequencies and levels of allelic diversity to evaluate whether local enhancement activities have had any effect on genetic diversity or population structure. These objectives were accomplished by analyzing varia- tion at 14 microsatellite loci to evaluate relationships among Pacific Rim populations of chum salmon. The distribution of genetic diversity among regions, popula- tions, and sampling years was estimated in the study. Materials and methods More than 53,000 chum salmon from 381 populations from Korea, Japan, Russia, Alaska, Canada, and Wash- ington were analyed from 59 geographic regions (Table 1, Fig. 1), with the specific populations and sample sizes outlined by Beacham et al.1 Tissue samples were collected from mature chum salmon, preserved in 95% ethanol, and analyzed at the Molecular Genetics Labo- ratory at the Pacific Biological Station (Fisheries and Oceans Canada, Nanaimo, BC). DNA was extracted from the tissue samples using a variety of methods, including a chelex resin protocol outlined by Small et al. (1998), a Qiagen 96-well Dneasy® procedure (Mis- sissauga, Ontario), or a Promega Wizard SV96 Genomic DNA Purification system (Promega, Madison, WI). Once DNA was extracted, surveys of variation at 14 micro- satellite loci were conducted: Ots3 (Banks et al., 1999), Oke3 (Buchholz et al., 2001), Oki2 (Smith et al., 1998), OkilOO (Beacham et al., 2008a), Omml070 (Rexroad et al., 2001), OmylOll (Spies et al., 2005), OnelOl, Onel02, Onel04, Onelll, and OtielM (Olsen et al., 2000), Otsl03 (Nelson and Beacham, 1999), Ssa419 (Cairney et al., 2000), and OtsG68 (Williamson et al., 2002). In general, polymerase chain reaction (PCR) DNA amplifications were conducted using DNA Engine Cycler Tetrad2 (BioRad, Hercules, CA) in 6,uL volumes consist- ing of 0.15 units of Taq polymerase, 1 pL of extracted 1 Beacham, T. D., J. R. Candy, S. Urawa, S. Sato, N. V. Var- navskaya, K. D. Le, and M. Wetklo. 2008. Microsatellite stock identification of chum salmon on a Pacific Rim basis and a comparison with single nucleotide polymorphisms (SNPs). Manuscript in review. DNA, lx PCR Hotstar buffer (Qiagen, Mississauga, Ontario, Canada), 60 pM each nucleotide, 0.40 pM of each primer, and deionized water. The thermal cycling profile involved one cycle of 15 minutes at 95°C, fol- lowed by 30-40 cycles of 20 seconds at 94°C, 30 to 60 seconds at 47-65°C and 30 to 60 seconds at 68-72°C (depending on the locus). Specific PCR conditions for a particular locus could vary from this general summary as outlined by Beacham et al. (in press). PCR fragments were initially size fractionated in denaturing polyacryl- amide gels using an ABI 377 automated DNA sequencer (Applied Biosystems, Foster City, CA), and genotypes were scored by Genotyper 2.5 software (Applied Bio- systems, Foster City, CA) using an internal lane sizing standard. Later in the study, microsatellites were size fractionated in an ABI 3730 capillary DNA sequencer (Applied Biosystems, Foster City, CA), and genotypes were scored by GeneMapper software 3.0 (Applied Bio- systems, Foster City, CA) using an internal lane sizing standard. Allele identification between the two sequenc- ers were standardized by analyzing approximately 600 individuals on both platforms and converting the sizing in the gel-based data set to match that obtained from the capillary-based set. Data analysis All annual samples available for a location were com- bined to estimate population allele frequencies, as was recommended by Waples (1990). Each population at each locus was tested for departure from Hardy-Weinberg equilibrium by using the computer software Genetic Data Analysis (GDA) (Univ. of Connecticut, Storrs, CT). Critical significance levels for simultaneous tests were evaluated using sequential Bonferroni adjustment (Rice 1989). Weir and Cockerham’s (1984) FST estimates for each locus over all populations were calculated with FSTAT version 2. 9. 3. 2 (Goudet, 1995). The significance of the multilocus FST value over all samples was deter- mined by jackknifing the FgT value over loci. The 59 geographic regions outlined in Table 1 were combined into 15 larger regional groups as outlined in Table 3 in order to display mean pairwise FST values between regions, but the two additional continental reporting groups (Asia, North America) incorporated in Table 3 were not used in the analysis of regional FST variation. Cavalli-Sforza and Edwards (CSE) (1967) chord dis- tance was used to estimate genetic distances among all populations. An unrooted neighbor-joining tree based upon CSE was generated using NJPLOT (Perriere and Gouy, 1996). Bootstrap support (by sampling loci) for the major nodes in the tree was evaluated with the CONSENSE program in PHYLIP software, based upon 1000 replicate trees (Felsenstein, 1993). FSTAT was used to measure the “allelic richness” (allelic diver- sity standardized to a sample size of 911 fish) for each regional group of populations evaluated. The distribu- tion of genetic variation in chum salmon was evaluated among regions, among populations within regions, and among sampling years within populations. In order to 246 Fishery Bulletin 107(2) Figure I Map of the Pacific Rim indicating the general geographic regions where chum salmon (Oncorhynchus keta ) from 381 populations were surveyed. The regions are listed in Table 1. Beacham et al. : Population structure of Oncorhynchus keta across the Pacific Rim 247 Table 1 Summary of the number of sampling sites or populations of chum salmon (Oncorhynchus keta) within each geographic region listed in Figure 1. A complete listing of the populations is outlined by Beacham et al.1 in their Appendix Table 1. n is the number of populations sampled within regions. The range of population sample sizes within regions is given in parentheses. Geographic area Reporting region n Mean population sample size Korea Korea i 100 (100-100) Japan Honshu Island, Sea of Japan Coast 5 106(80-160) Hokkaido Island, Sea of Japan Coast 3 147 (60-280) Hokkaido Island, Sea of Okhotsk Coast 5 108(50-160) Hokkaido Island, Nemuro Strait 2 95 (80-110) Hokkaido Island, eastern Pacific Coast 2 105(80-130) Hokkaido Island, western Pacific Coast 4 120(80-160) Honshu Island, Pacific Coast 5 68(19-80) Russia Primorye 3 34(17-49) Amur River 1 338(338-338) Sakhalin Island 4 76 (49-149) Magadan 5 89 (55-120) Northern Sea of Okhotsk 2 60 (43-76) West Kamchatka 8 116 (40-249) East Kamchatka 9 58(39-128) Northeast Russia 2 87(79-94) Arctic Canada Mackenzie River 1 33(33-33) Yukon River Lower river summer run (United States) 11 185 (92-347) Tanana River summer run (United States) 2 211 (185-236) Tanana River fall run (United States) 3 160(80-241) Upper Alaska (United States) 4 149 (73-229) Porcupine River (Canada) 2 463 (329-597) White River (Canada) 3 207(62-486) Mainstem Yukon River (Canada) 4 144(83-175) Teslin River (Canada) 1 143(143-143) Upper Yukon River early fall (Canada) 1 120 (120-120) Western Alaska Kotzebue Sound 6 155 (45-374) Norton Sound 10 278(50-474) Kuskokwim River and bay 6 94(68-171) Nushagak River 2 78(74-82) North Central Bristol Bay 4 77 (64-92) South Bristol Bay 4 83 (57-97) North Peninsula and Aleutians 3 122 (93-179) Central Alaska Southwest Peninsula 4 83(70-104) Southeast Peninsula 3 91 (87-94) Kodiak Island 3 89 (71-100) Prince William Sound 4 98 (92-100) Southeast Alaska Southeast Alaska 14 119 (50-333) Queen Charlotte Islands West Coast 11 209(42-393) North Coast 4 132 (80-221) East Coast 11 161 (17-376) Skidegate Channel 8 181 (79-232) Continued 248 Fishery Bulletin 107(2) Table 1 (continued) Geographic area Reporting region n Mean population sample size Northern British Columbia Taku River 5 34(12-65) North Coast 18 117(28-242) Skeena River 13 95(12-182) Grenville Channel 6 122 (40-242) Central Coast 52 190 (13-419) Rivers Inlet 8 79 (40-153) Smith Inlet 2 363 (226-499) Southern British Columbia Johnstone Strait 13 134(20-409) South Coast 14 137(15-344) Vancouver Island east coast 9 227 (167-285) Vancouver Island west coast 10 133 (43-243) Fraser River 23 151 (24-427) Washington North Puget Sound 7 85 (50-100) South Puget Sound 3 100(100-100) Hood Canal 2 95 (88-102) Strait of Juan de Fuca 2 100(100-100) Coast of Washington 4 91 (61-106) maintain a balanced design, regions included in the analysis required two or more populations each with two or more years of samples available. Regions were distributed around the Pacific Rim and were a subset of the 59 geographic regions outlined in Table 1 and Figure 1. The specific populations included from each region are in shown parentheses: West Kamchatka (Hairusova, Vorovskaya), Western Alaska (Snake, Eldorado), Yukon River summer run (Gisasa, Tozitna), Southeast Alaska (DIPAC hatchery, Disappearance), Queen Charlotte Islands west coast (Clapp Basin, Mace), Queen Charlotte Islands east coast (Lagoon, Pallant), Northern Brit- ish Columbia (Ensheshese, Kateen), Grenville Channel (Markle, Wilson), British Columbia central coast (Bull- ock Channel, Quaal, Salmon), Smith Inlet (Walkum, Nekite), Johnstone Strait (Viner Sound, Nimpkish), Vancouver Island east coast (Big Qualicum, Cowichan), and Fraser River (Inch, Stave). Estimation of variance components of river drainage or region differentiation, among populations within drainages or regions, and among years within populations was determined with Genetic Data Analysis. Results Variation within populations Substantial variation was observed in the number of alleles at the 14 microsatellite loci surveyed in the study. The fewest number of alleles was observed at Oke3 (26 alleles), and the greatest number of alleles observed at Onelll (149 alleles) (Table 2). Lower heterozygosity was observed at loci with fewer than 40 alleles. The genotypic frequencies at each locus conformed to those expected under Hardy-Weinberg equilibrium (HWE). The number of alleles observed displayed considerable variation across regional groups of chum salmon. Asian chum salmon were considerably more diverse than those in North America, with Asian populations displaying the greatest number of alleles at all 14 loci examined (P=0.0001) (Table 3). With sample sizes standardized to 911 fish per region, Japanese chum salmon were the most genetically diverse set of populations examined with 581 alleles observed, greater than in all other re- gional groups of populations. The least diverse groups of populations were observed in the Queen Charlotte Islands, the Skeena River, the east coast of Vancouver Island, and Washington State, with an average of 414 alleles observed in chum salmon from these regions. Japanese chum salmon displayed 40% more alleles and Russian chum salmon 35% more alleles than did chum salmon from the four regions of lower genetic diversity. The greatest difference in diversity was observed at locus Onelll, with the greatest number of observed alleles, and Asian chum salmon displayed 80% more alleles than did chum salmon from the four regions of lower genetic diversity. Even with Onelll removed from the analysis, Asian chum salmon were still more diverse than chum salmon in all regions in North America (P=0.0002). Distribution of genetic variance Gene diversity analysis of the 14 loci surveyed was used to evaluate the distribution of genetic variation Beacham et al.: Population structure of Oncorhynchus keta across the Pacific Rim 249 Table 2 Number of alleles per locus, an index of gentic differentiation FST (SD in parentheses), expected heterozygosity (He), observed heterozygosity (Hn), and percent significant Hardy-Weinberg equilibrium (HWE) test for 14 microsatellites ( m = 381 tests) among 381 chum salmon (Oncorhynchus keta) populations. Locus Number of alleles ^ST He H0 HWE 1 Oke3 26 0.104(0.005) 0.67 0.65 3.7 2 OkilOO 31 0.039 (0.002) 0.83 0.83 0.3 3 Ots3 31 0.097 (0.005) 0.76 0.75 4.7 4 Oki2 42 0.062(0.005) 0.86 0.85 0.8 5 OmylOll 44 0.027(0.001) 0.90 0.89 0.3 6 One 104 48 0.027 (0.001) 0.93 0.92 1.6 7 Otsl03 54 0.019 (0.001) 0.94 0.93 1.1 8 Ssa419 54 0.028(0.001) 0.84 0.83 0.5 9 One 101 56 0.058 (0.002) 0.87 0.86 1.6 10 Omml070 60 0.009 (0.000) 0.95 0.94 1.3 11 One 114 60 0.017(0.001) 0.92 0.91 1.6 12 Onel02 69 0.011 (0.001) 0.92 0.90 2.1 13 OtsG68 69 0.017 (0.001) 0.95 0.94 1.8 14 Onelll 149 0.036 (0.003) 0.94 0.93 3.9 Total 0.033(0.007) 0.88 0.87 Table 3 Mean number of alleles observed per locus at 14 microsatellite loci for chum salmon (Oncorhynchus keta) from 15 geographic areas standardized to a sample size of 911 fish per geographic area. Geographic areas, listed in Table 1, were: Japan (includes Korea), Russia, Western Alaska ( WAK), Yukon River (includes Arctic Canada), Central Alaska (CAK), Southeast Alaska (SeAK), Queen Charlotte Islands (QCI), northern British Columbia (NBC), Skeena River, Central Coast British Columbia (CBC) (includes Grenville Channel, Rivers Inlet, and Smith Inlet), Southern British Columbia (includes Johnstone Strait), east coast Vancouver Island (ECVI), west coast Vancouver Island ( WCVI), Fraser River, Washington (Wash), and North America (NA). Area Oke 3 Oki 100 Oki 2 Omm 1070 Omy 1011 One 101 One 102 One 104 One 111 One 114 Ots 3 Ots 103 Ots G68 Ssa 419 Total Japan 16.6 25.0 23.8 51.0 39.5 37.8 42.3 37.4 122.0 34.5 25.8 46.8 53.1 25.2 580.8 Russia 14.9 27.1 18.9 45.1 31.3 40.2 28.4 34.4 130.2 40.6 22.4 47.6 51.2 26.6 558.9 Total Asia 15.8 26.1 21.4 48.1 35.4 39.0 35.4 35.9 126.1 37.6 24.1 47.2 52.2 25.9 569.9 WAK 9.9 24.2 18.2 37.2 28.1 33.0 21.7 27.7 110.7 39.9 19.2 40.4 41.7 17.9 469.8 Yukon R. 12.9 22.8 20.5 37.7 28.6 35.9 28.2 29.3 112.1 35.0 21.2 38.1 41.3 16.7 480.3 CAK 9.7 25.2 20.0 36.3 25.7 29.5 27.0 30.6 106.4 35.4 18.9 39.7 40.6 20.2 465.2 SeAK 8.6 21.6 20.4 40.9 24.3 37.6 28.9 28.5 94.7 35.5 18.6 43.3 43.1 22.8 468.8 QCI 11.8 17.2 20.8 38.3 21.1 36.6 29.1 28.5 69.7 26.1 18.1 39.7 44.0 23.1 424.1 NBC 14.5 19.2 20.3 40.4 26.5 41.0 31.3 31.3 102.0 31.1 18.2 42.5 45.8 25.2 489.3 Skeena R. 7.5 15.8 18.0 37.3 22.7 35.9 25.9 28.0 65.3 25.2 16.1 37.5 42.0 17.7 394.9 CBC 13.9 19.3 20.2 41.7 26.2 39.0 31.4 28.6 93.8 33.7 21.4 42.1 45.2 22.7 479.2 SBC 16.2 17.9 20.2 38.8 26.0 40.6 28.8 33.0 76.1 31.2 22.0 39.1 44.8 20.5 455.2 ECVI 8.0 15.7 23.0 39.8 21.8 37.0 24.9 25.9 73.9 27.0 19.6 36.9 44.9 19.0 417.4 WCVI 11.0 17.8 25.8 35.9 24.7 39.8 28.2 29.8 74.6 31.7 22.4 39.0 47.4 16.8 444.9 Fraser R. 13.0 20.8 18.1 42.2 25.0 38.8 23.4 31.1 89.6 34.3 20.6 38.7 56.0 15.4 467.0 Washington 10.5 17.6 18.0 38.8 18.8 36.3 26.6 29.0 71.0 34.0 15.8 40.5 45.5 15.6 418.0 Total NA 11.3 19.6 20.3 38.9 24.6 37.0 27.3 29.3 87.7 32.3 18.2 39.8 44.8 19.5 451.8 250 Fishery Bulletin 107(2) Table 4 Hierarchical gene-diversity analysis of 27 populations of chum salmon ( Oncorhynchus keta) within 13 regions for 14 microsatel- lite loci. Regions had a Pacific Rim distribution, and the time difference between the earliest and latest samples included for specific populations ranged from 1 to 21 years. Ratio is the sum of the variance components of among populations within regions and among regions divided by the variance component among years within populations. * P<0.05 ** P<0.01. Locus Within populations Among years within populations Among populations within regions Among regions Ratio Oke3 0.9204 0.0004 0.0056** 0.0736** 198.0 OkilOO 0.9673 0.0008 0.0070** 0.0249** 39.9 Ots3 0.9254 0.0018* 0.0044** 0.0685** 40.5 Oki2 0.9625 0.0065** 0.0044* 0.0266** 4.8 OmylOll 0.9783 0.0016 0.0013 0.0187** 12.5 Onel04 0.9783 0.0004 0.0030** 0.0183** 53.3 Otsl03 0.9837 0.0007 0.0029** 0.0126** 22.1 Ssa419 0.9785 0.0018* 0.0054** 0.0143** 10.9 OnelOl 0.9635 0.0015 0.0088** 0.0262** 23.3 Omml070 0.9918 0.0011 0.0031** 0.0040* 6.5 Onell4 0.9881 0.0009 0.0042** 0.0068* 12.2 Onel02 0.9941 0.0008 0.0018 0.0033* 6.4 OtsG68 0.9854 0.0015 0.0036** 0.0095** 8.7 Onelll 0.9727 0.0016* 0.0030 0.0227** 16.1 Total 0.9722 0.0015 0.0041** 0.0222** 17.5 among regions, among populations within regions, and among years within populations. Within popula- tions, the time difference between the earliest and latest samples included in the analysis ranged from 21 years (1986-2007) for Disappearance Creek (Southeast Alaska), 18 years (1986-2004) for Lagoon Creek (Queen Charlotte Islands), 16 years (1989-2005) for Nekite River in Smith Inlet, 15 years (1988-2003) for Gisasa River (Lower Yukon River), down to 1-3 year differences for populations in a number of regions. For 13 regions ranging from west Kamchatka to the Fraser River, the amount of variation within populations ranged from 92% ( Oke3 ) to 99% (Omml070), with the average for an individual locus 97% (Table 4). Variation among the 13 regions included in the analysis accounted for 2.2% of total observed variation. Variation among populations within regions accounted for 0.4% of observed variation, with differences among regions over five times greater than differences among populations within regions. The variation among sampling years within populations was the smallest source of variation observed, accounting for 0.2% of all variation. Differentiation among regions and populations within regions was approximately 18 times greater than that of annual variation within populations. For the time intervals surveyed in our study, annual variation in microsatellite allele frequen- cies was relatively minor compared with differences among populations within regions and among regions on a geographically diverse scale of distribution of the populations analyzed. Population structure Significant genetic differentiation was observed among chum salmon populations sampled in the different geographic regions surveyed. The FST value over all populations and loci was 0.033, with individual locus values ranging from 0.009 ( Omm 1070 ) to 0.104 ( Oke3 ) (Table 2). Chum salmon from Japan and the Yukon River were among the most distinct regional groups of stocks included in the survey (Table 5). Greatest genetic differentiation (greatest difference in FST values) was observed in comparisons between Japanese, Russian, western Alaskan, and Yukon River chum salmon com- pared with those in other regions in North America to the south and east. In Asia, chum salmon from Japan were generally distinct from those in Russia. In North America, significant regional differentiation was gen- erally observed, with chum salmon in more northern regions distinct from those in more southern regions. Two major lineages of chum salmon populations were identified in the cluster analysis. The first lineage in- cluded all populations sampled from Korea, Japan, Russia, the Mackenzie River, Kotzebue Sound, Norton Sound, the Yukon River, and northern and central Bris- tol Bay. Populations from southern Bristol Bay were intermediate between the two major lineages, and all populations from the Alaska Peninsula south and east to Washington State were identified as the second ma- jor lineage (Fig. 2). Within the first lineage, all Asian populations were distinct from all North American Beacham et al.: Population structure of Oncorhynchus keta across the Pacific Rim 251 Table 5 Mean pairwise FST values averaged over 14 microsatellite loci from 15 regional groups of chum salmon ( Oncorhynchus keta) outlined in Table 3 that were sampled at 381 locations across the Pacific Rim. Comparisons were conducted between individual populations in each region. Values in bold are the diagonal, and are comparisons among populations within each region. FST values are listed below the diagonal, with standard deviations above the diagonal. Some of the reporting regions listed in Table 1 were combined as indicated in Table 3 in order to facilitate the analysis. RC is region code, and codes are as follows: 1) Japan, 2) Russia, 3) Western Alaska, 4) Yukon River, 5) Central Alaska, 6) Southeast Alaska, 7) Queen Charlotte Islands, 8) Northern British Columbia, 9) Skeena River, 10) Central British Columbia, 11) Southern mainland British Columbia, 12) West coast Van- couver Island, 13) East coast Vancouver Island, 14) Fraser River, 15) Washington. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 0.019 0.009 0.014 0.023 0.015 0.007 0.008 0.009 0.009 0.009 0.008 0.007 0.008 0.011 0.011 2 0.026 0.017 0.013 0.018 0.016 0.011 0.010 0.011 0.012 0.011 0.011 0.010 0.010 0.011 0.014 3 0.028 0.024 0.012 0.018 0.018 0.011 0.011 0.011 0.012 0.011 0.009 0.010 0.011 0.015 0.013 4 0.053 0.054 0.031 0.018 0.020 0.013 0.014 0.014 0.019 0.014 0.015 0.013 0.017 0.022 0.016 5 0.042 0.032 0.037 0.064 0.027 0.014 0.016 0.015 0.019 0.016 0.014 0.015 0.013 0.011 0.015 6 0.042 0.029 0.035 0.062 0.024 0.007 0.006 0.005 0.016 0.006 0.007 0.005 0.011 0.006 0.009 7 0.050 0.039 0.043 0.068 0.034 0.015 0.012 0.007 0.017 0.007 0.007 0.005 0.014 0.008 0.010 8 0.044 0.031 0.037 0.063 0.026 0.008 0.015 0.008 0.017 0.007 0.009 0.007 0.013 0.009 0.012 9 0.053 0.041 0.046 0.066 0.035 0.019 0.025 0.019 0.014 0.017 0.014 0.012 0.017 0.015 0.017 10 0.043 0.031 0.037 0.062 0.030 0.011 0.014 0.010 0.020 0.008 0.007 0.005 0.013 0.008 0.009 11 0.046 0.033 0.040 0.068 0.039 0.022 0.025 0.021 0.030 0.018 0.014 0.007 0.019 0.012 0.012 12 0.044 0.034 0.038 0.062 0.038 0.019 0.018 0.019 0.026 0.017 0.016 0.008 0.016 0.009 0.010 13 0.043 0.032 0.034 0.060 0.039 0.026 0.031 0.026 0.035 0.025 0.019 0.022 0.022 0.018 0.011 14 0.041 0.028 0.035 0.063 0.037 0.025 0.030 0.026 0.033 0.024 0.018 0.021 0.020 0.013 0.015 15 0.051 0.039 0.047 0.076 0.045 0.028 0.033 0.029 0.035 0.025 0.022 0.023 0.028 0.022 0.022 populations. Within the Asian portion of the lineage, Japanese, Korean, and Russian Primorye populations were distinct from other Asian populations. In the sec- ond lineage, populations from Washington and southern British Columbia were among the most distinct group of populations, along with populations from the Queen Charlotte Islands in northern British Columbia. Chum salmon spawning in tributaries of different major river drainages generally clustered together in the analysis. For example, Fraser River populations clustered together in 39% of dendrograms evaluated, Skeena River populations clustered together in 97% of dendrograms evaluated, and Taku River populations clustered together in 96% of dendrograms evaluated (Fig. 2). The one exception was the Yukon River, where lower river summer-run populations did not form dis- tinct clusters unique from neighboring populations in the Kuskokwim River and the Nushagak River. A very distinct regional cluster of populations was observed in the Asian populations, with Korean, Japa- nese, and populations from the Primorye region in Rus- sia clustering together in 100% of dendrograms evalu- ated. Within that cluster, populations from Primorye clustered together in 67% of dendrograms evaluated, indicative of genetic differentiation between popula- tions from that region and those in Japan and Korea. Within Japan, a general regional structuring of popula- tions was observed, with populations from the Pacific coast of Honshu Island forming a distinct group (92% of dendrograms evaluated), as did populations from the Nemuro Strait (89% of dendrograms) and the eastern Pacific coast of Hokkaido Island (50% of dendrograms). Within Russia, Magadan region populations clustered together in 41% of dendrograms evaluated, as did pop- ulations from the northern Sea of Okhotsk (100% of dendrograms). Although populations from east coast of Kamchatka and west coast of Kamchatka generally clustered as two distinct regional groups, the groupings were not strongly supported by the bootstrap analysis. Populations from northeast Russia were distinct from those in other regions, with the possible exception of the Utka River population from west Kamchatka. In North America, some level of regional structuring of populations was observed in both Kotzebue Sound and Norton Sound (Fig. 2). Within the Yukon River drainage, there was clear separation between sum- mer-run populations in the lower and mid- portions of the drainage and fall-run populations in the upper portion of the drainage. For the fall-run, populations in the White River in the Yukon Territory were quite distinct, clustering together in 100% of dendrograms evaluated. Similarly, fall-run populations in the Ta- nana River (upper portion of Yukon River drainage in Alaska) clustered together in 74% of the dendrograms evaluated, and summer-run populations in the Tanana River drainage clustered together in 96% of dendro- grams. Summer-run populations in the lower Yukon River drainage did not cluster exclusively with each 252 Fishery Bulletin 107(2) Orikasa Namdae I Korea j 0.002 | Tsugaruishi | Hokkaido - eastern Pacific Hokkaido - western Pacific ~Ohkawa Honshu Pacific Uono Hayatsuki Honshu - Sea of Japan Toshibetsu Hokkaido - Sea of Japan Horonai Hokkaido - Sea of Okhotsk Nishibetsu l Hokkaido - Nemuro Strait Tokushibetsu I Hokkaido - Sea of Okhotsk Tcshio i Hokkaido - Sea of Japan Avakumovka 100 ■ Udarnitsa Ryazanovka ■ Tugur Naiba " lym Narva | Primorye Kalininka Sakhalin Island Amur i Amur -Kol "Kikchik Plotnikova Bolshaya "Okhota "Hairusova Pymta ' Vorovskaya West Kamchatka ■ Zhypanova " Kamchatka — Ulutorsky Apuka Neipichi ■ Ossora ~ Dranka • Ivashka Karaga East Kamchatka Ola L 41 - Magadan Tauy 100 Magadan Penzhina ■ Oklan 100 L Kanchalan Anadyr Northern Sea of Okhotsk Utka Northeast Russia North / Central Bristol Bay Peel I Mackenzie River Kotzebue Sound Lower Yukon summer / Kuskokwim Lower Yukon summer / Kuskokwim Figure 2 Neighbour-joining dendrogram of Cavalli-Sforza and Edwards (1967) chord distance for 381 Pacific Rim populations of chum salmon (Oncorhynchus keta) surveyed at 14 microsatellite loci. Bootstrap values at major tree nodes indicate the percentage of 1000 trees where populations beyond the node clustered together. Beacham et al.: Population structure of Oncorhynchus keta across the Pacific Rim 253 other, including populations from the Kuskokwim River in western Alaska and Nushagak River from northern Bristol Bay in the dendrogram cluster. Geographically-based regional clustering was observed in the populations surveyed south and east of north- ern Bristol Bay. Populations from southern Bristol Bay formed a distinct cluster in 98% of dendrograms evalu- ated, with bootstrap support observed for populations from the western south coast of the Alaska Peninsula, eastern south coast of the Alaska Peninsula, Kodiak 254 Fishery Bulletin 107(2) Grenville Channel Taaltz Johnstone Strait Central Washington Figure 2 (continued) Island, and Prince William Sound. Populations from northern southeast Alaska formed a distinct cluster in the analysis, but populations from southern southeast Alaska were less distinct than those in the northern portion of the region. Some clusters in the dendrogram included populations from both southern southeast Alas- ka and northern British Columbia (Fig. 2). In British Columbia (BC), four geographically based regional groups of populations were observed in the Queen Charlotte Islands (QCI). North coast QCI popu- Beacham et al. : Population structure of Oncorhynchus keta across the Pacific Rim 255 Fraser River South Puget Sound Hood Canal Hood Canal / Strait of Juan de Fuca south coast Vancouver Island - east coast ■ Ensheshese BC - north coast Carroll I Southeast Alaska BC - north coast " Kumealon ‘ Klewnuggit Nakut i Southeast Alaska “ Lachmach Stumaun Crag Lagoon Lizard BC - north coast Neets Bay early Disappearance “Neets Bay late Fish llliance Wilauks Southeast Alaska BC - north coast Upper Kitsumkalum Nangeese Skeena River Figure 2 (continued) lations were the most distinct clustering together in 100% of dendrograms evaluated. Regional populations were also observed along the east and west coasts of the QCI. Populations adjacent to Skidegate Channel, the body of water separating the major QCI compo- nents of Graham Island (north) and Moresby Island (south), clustered together with 64% bootstrap support. On the northern mainland, populations north of the Skeena River mouth were distinct from those south of the Skeena River. North of the Skeena River, there 256 Fishery Bulletin 107(2) were not distinct clusters observed between northern coastal British Columbia populations and those from southeast Alaska. Populations immediately south of the Skeena River in the Grenville Channel area clus- tered separately from those further south in the central coastal region of British Columbia. Yet farther south, populations from Rivers Inlet and Smith Inlet clustered together in geographically based groups, and this result was confirmed by 100% boostrap support observed for Smith Inlet populations (Fig. 2). In southern BC, five geographically based groups of populations were revealed. East coast and west coast of Vancouver Island populations were regionally separate from each other, and also from other regional popu- lations in southern BC. On the mainland, Johnstone Strait populations were separate from those in southern coastal areas, and the demarcation point between the two groups is Bute Inlet, at the northeast limit of the Strait of Georgia. Fraser River populations were dis- tinct from other regional groups in southern BC. In Washington, regional structuring of chum popula- tions was observed. The most distinct regional group comprised populations from the outer Pacific coast, with populations clustering together with 100% boot- strap support. In more inside waters, populations from north Puget Sound were generally distinct from those in south Puget Sound, Hood Canal, and the Strait of Juan de Fuca. South Puget Sound populations were distinct from those in Hood Canal and the Strait of Juan de Fuca. Discussion The survey of microsatellite variation included an exami- nation of variation at 14 loci encompassing approximately 800 alleles, with 26 to 149 alleles recognized per locus. The number of fish surveyed per population ranged from 12 to 597 individuals (Beacham et al.1). With a vari- able number of individuals surveyed per population, there was a potential for sampling error in estimated allele frequencies and in obscuring genetic relation- ships among related populations, particularly if sample sizes were small for some populations in a lineage. For example, for the Primorye populations from Russia, pop- ulation sample sizes ranged from 17 to 49 individuals, and it was possible that estimates of genetic distances among populations were not determined satisfactorily for populations of smaller sample size, particularly for those loci with larger numbers of alleles. However, Kalinowski (2005) reported that loci with larger numbers of alleles (higher mutation rates) produced estimates of genetic distance with lower coefficients of variation than loci with fewer numbers of alleles, without requiring larger sample sizes from each population. Population structur- ing based upon geographic differences were observed for populations from Primorye, and all populations clustered together in 67% of dendrograms evaluated. Therefore, it seems likely that variation in the number of individuals surveyed within a population in our study did not gener- ally result in misidentification of genetic relationships among populations. Size homoplasy of microsatellite alleles may have some effect on the estimate of genetic differentiation ob- served among populations. Inferences about the genetic relationships of populations surveyed in our study were dependent upon accurate determination of population allele frequencies. Microsatellite alleles differ in size, but alleles of the same size at a locus in geographically disparate populations may not have the same origin as a result of size homoplasy. Convergent mutations in different lineages may produce alleles of the same size, with the result that there may be greater differentiation among lineages than revealed by analysis of size varia- tion. However, with approximately 800 alleles observed across all loci in the study, the large amount of varia- tion present at these loci largely compensates for size homoplasy (Estoup et al., 2002). In this study, population allele frequencies were es- timated by combining all samples collected over time for a population, regardless of the length of time that occurred between samples. In practice, the maximum length of time between samples for a population was 21 years, and up to six annual samples were combined for a population. Analysis of the distribution of genetic variation indicated that differentiation among regions and populations within regions was approximately 18 times greater than that of annual variation within populations, indicating that pooling of annual samples over time is a practical approach to estimate population allele frequencies. Relative stability of microsatellite al- lele frequencies over time is not unique to chum salmon; similar relative stability has been reported for sockeye salmon (O. nerka) (Beacham et al., 2006a) and Chinook salmon (O. tshawytscha ) (Beacham et al., 2006b). Surveys of genetically based population structure in chum salmon were initially conducted with allo- zymes. Okazaki (1982b), in a study evaluating allozyme variation in Asian and North American populations, concluded that there were 11 geographically based re- gional groups of populations across the Pacific Rim. The regional groups consisted of adjacent river populations that were genetically similar within one region. Many allozyme-based studies of regional population structure were subsequently reported. For example, Winans et al. ( 1994) provided additional details concerning population structure of Asian populations, Wilmot et al. (1994) compared population structure of chum salmon from western Alaska and northeast Russia, Kondzela et al. (1994) compared population structure of chum salmon from southeast Alaska and northern British Columbia, Beacham et al.( 1987) evaluated population structure of chum salmon in British Columbia, and Phelps et al. (1994) evaluated population structure in the Pacific Northwest. Seeb and Crane (1999) again investigated chum salmon population structure throughout the Pa- cific Rim by examining variation at 40 allozymes, and reported that two major lineages of populations were observed. The northern lineage occurred in areas north of the Alaska Peninsula and into Russia and Japan, Beacham et al. : Population structure of Oncorhynchus keta across the Pacific Rim 257 whereas the southern lineage was observed in the Alas- ka Peninsula, Kodiak Island, and areas to the south and east. The two lineages were reported to overlap in the northern Alaska Peninsula. Development of DNA-level markers provided addition- al markers for genetic evaluation of population struc- ture of chum salmon, and surveys of mitochondrial DNA variation have been reported. Differentiation among Russian populations has been reported (Ginatulina, 1992; Brykov, 2003; Polyakova et al., 2006), as well as in Japanese populations (Sato et al., 2001). In an analy- sis of mtDNA variation across the Pacific Rim, Sato et al. (2004) reported that there were three major lineages of chum salmon, with populations from Japan, Russia, and North America comprising three distinct regional groups. Chum salmon from Japan were observed to be the most distinct, with less divergence between popula- tions from Russia and western Alaska. Minisatellite variation was used by Taylor et al. (1994) and Beacham (1996) to survey variation in 42 chum salmon populations across the Pacific Rim. Three regional groups of populations showed that those from Japan were the most distinct, followed by a second (less distinct) group comprising Russian and Yukon River populations, and a third group comprising southeast Alaska and British Columbia populations. Microsatellites have been used to evaluate chum salmon population differentiation and structure on a local and regional basis (Chen et al., 2005; Beacham et al., 2008a, 2008b, 2008c, in press). In those studies, as in the previous allozyme-based studies, regional groups of populations were observed, with the regional groups consisting of adjacent river populations or local popula- tions that were genetically similar within one region. The results from the current study were remarkably similar to the results of the allozyme-based study re- ported by Seeb and Crane (1999), with populations from Korea, Japan, Russia, Kotzebue Sound, Norton Sound, the Yukon River, and northern Bristol Bay determined to be in one major lineage. Populations from southern Bristol Bay and the northern Alaska Peninsula were intermediate, and populations on the south side of the Alaska Peninsula, Kodiak Island, and areas to the south and east to Washington State were determined to be in a second major lineage. Successful transplantation of salmon within the range of a species has the potential to alter genetic popula- tion structure. Population structure of chum salmon has been influenced to some degree by transplantations within its range. For example, due to frequent trans- plantations associated with hatcheries, most Japanese populations have received some level of transplantation of non-natal fish. Although initial studies indicated that the effect of transplantations were minimal in Japanese populations (Okazaki, 1982a), more recent work has shown that some current run-timing variation in popu- lations may be a result of transplantations. Beacham et al. (2008b) reported that allozyme monitoring indi- cated that successful introduction and establishment of broodstock from the Chitose River on the Sea of Japan coast of Hokkaido Island to the Gakko River on the Sea of Japan coast of Honshu Island accounts for observed temporal differentiation in the existing Gakko River population. Transplantations have also occurred in Russian and North American populations, but there is little evidence for a demonstrable change in popula- tion structure as a result of transplantations. Although most production of Japanese chum salmon is currently derived from hatcheries, there is little evi- dence that hatchery production has resulted in reduced genetic variation of the populations, in relation to chum salmon in other portions of the range. Initially, Kaeri- yama (1999) indicated that, on the basis of allozymes, Japanese populations were less variable than Russian wild populations. In our study, on the basis of 14 mic- rosatellites, we found no evidence that Japanese chum salmon populations were less genetically variable than Russian or North American chum salmon. In fact, the opposite result was observed, with higher levels of ge- netic variation observed in Japanese populations com- pared with chum salmon from other regions across the Pacific Rim. Population structure of chum salmon across the Pa- cific Rim was demonstrated to have a regional basis. A regionally based population structure is generally required for genetic stock identification estimation be- cause an important assumption is that the portion of the mixed-stock sample derived from unsampled popula- tions is allocated to sampled populations from the same region. This assumption reduces the cost and complexity of developing a baseline for stock composition analysis. Chum salmon population structure thus meets the im- portant condition that unsampled populations contribut- ing to mixed fishery samples will likely be allocated to sampled populations in the same region. Populations in the major river drainages surveyed all clustered together within a drainage, with the excep- tion of the Yukon River, where lower river summer-run populations clustered with populations from the Kus- kokwim River in western Alaska and the Nushagak River in northern Bristol Bay. Similar results were also reported in the allozyme survey conducted by Seeb and Crane (1999), who suggested that genetic exchange may have occurred between the Kuskokwim and Nushagak rivers during the last glaciation because both rivers were headwaters to a Bering Sea Land Bridge river that drained into the Bering Sea (Hopkins, 1967; Lindsay and McPhail, 1986). The ancient mouth of the Yukon River was farther south than at present (Hopkins, 1967; Knebel and Creager, 1973), increasing the probability of genetic exchange among ancestral populations of the Yukon, Kuskokwim, and Nushagak rivers. Chum salmon likely had a different pattern of dis- persal from refuges after the last glaciation ended in the Pleistocene Era some 10,000 years ago than did either sockeye salmon or Chinook salmon. For exam- ple, evaluation of genetic diversity in Asian and North American populations of sockeye salmon and Chinook salmon have indicated that there were similar levels of genetic diversity between populations from these 258 Fishery Bulletin 107(2) two continents (Beacham et al., 2006a, 2006b). This is in marked contrast to the pattern observed in chum salmon, with Asian chum salmon displaying signifi- cantly greater genetic diversity than that observed in chum salmon populations in North America. Surveys of mtDNA variation have also indicated that Japanese populations have the highest genetic diversity among Pacific Rim chum salmon (Sato et al., 2004). Chum salmon in Asia display a wider geographic distribution than either sockeye salmon or Chinook salmon, with most populations of these two species restricted to a Russian distribution, whereas chum salmon range as far south as South Korea. The distinctive nature of Korean, Japanese, and Primorye chum salmon, coupled with the higher diversity observed in Asian popula- tions, indicates an Asian refuge from which chum salm- on dispersed after the retreat of glaciers during the Pleistocene, either on the southern Asian mainland or the islands of Japan. The fact that Asian chum salmon display more genetic diversity than North American chum salmon reflects either that either higher popula- tion sizes were present in this refuge, allowing more genetic variation to be retained, or that dispersal from this refuge preceded those in North America, allowing more time for genetic mutations to accumulate. The concept of a glacial refuge near Japan was also sug- gested by Taylor et al. (1994). In North America, the observed population structure of chum salmon would support the concept at a mini- mum of a Bering Sea refuge in the north (unglaciated areas of western Alaska or Russia) and a Columbia River refuge in the south (unglaciated area west of the Continental Divide) as suggested by McPhail and Lindsey (1970). Present day populations in Korea, Ja- pan, and Primorye may be derived from the southern Asian (Japanese) refuge, populations from the Amur River through to southern Bristol Bay may be derived from the northern Bering refuge, and populations from the Alaska Peninsula to Washington may have been derived from the southern refuge. In British Columbia, an additional refuge may also have been present on the Queen Charlotte Islands (Warner et al., 1982). Queen Charlotte Islands chum salmon populations were dis- tinct and also displayed lower genetic variation, very similar to sockeye salmon populations from the region (Beacham et al., 2006a). Wood (1995) suggested that sockeye salmon population structure on the central coast region of British Columbia was consistent with colonization from two different refugia, and therefore it is possible that present day populations in British Columbia are derived from chum salmon originating from a Queen Charlotte Islands refuge and that other portions of the coast were colonized by chum salmon that originated from a southern refuge. Acknowledgments A very substantial effort was undertaken to obtain samples from chum salmon sampled in this study. In North America, starting from the south, we thank J. B. Shaklee, various staff of Fisheries and Oceans Canada (DFO) for baseline sample collection, as well as First Nations staff, R. L. Wilmot, L. W. Seeb, S. Johnston, P. Milligan, J. Wenburg, and A. J. Gharrett. For Asia, samples were provided by V. V. Efremov, N. V. Varnaks- kaya, G. Winans, S. Urawa, S. Sato, and J. Park. L. Fitzpatrick drafted the map. C. Wallace assisted in the analysis. Funding for the study was provided by Fisher- ies and Oceans, Canada. Literature cited Banks, M. A., M. S. Blouin, B. A. Baldwin, V. K. Rashbrook, H. A. Fitzgerald, S. M. Blankenship, and D. Hedgecock. 1999. Isolation and inheritance of novel microsatellites in chinook salmon ( Oncorhynchus tshawytscha). J. Hered. 90:281-288. Banks, M. A., V. K. Rashbrook, M. J. Calavetta, C. A. Dean, and D. Hedgecock. 2000. Analysis of microsatellite DNA resolves genetic structure and diversity of chinook salmon (Oncorhyn- chus tshawytscha ) in California’s Central Valley. Can. J. Fish. Aquat. Sci. 57:915-927. Beacham, T. D. 1996. The use of minisatellite DNA variation for stock identification of chum salmon, Oncorhynchus keta. Fish. Bull. 94:611-627. Beacham, T. D., A. R Gould, R. E. Withler, C. B. Murray, and L. W. Barner. 1987. Biochemical genetic survey and stock identifica- tion of chum salmon ( Oncorhynchus keta) in British Columbia. Can. J. Fish. Aquat. Sci. 44:1702-1713. Beacham, T. D., K. D. Le, M. Wetklo, B. McIntosh, T. Ming, and K. M. Miller. In press. Population structure and stock identification of chum salmon from western Alaska determined with microsatellite and major histocompatibility complex variation. In Pacific salmon: ecology and manage- ment of western Alaska’s populations (C. C. Krueger, and C. E. Zimmerman, eds.). Am. Fish. Soc., Symp., Bethesda, MD. Beacham, T. D., S. Urawa, K. D. Le, and M. Wetklo. 2008b. Population structure and stock identification of chum salmon from Japan determined with microsatel- lite DNA variation. Fish. Sci. 74:983-994. Beacham, T. D., B. Spilsted, K. D. Le, and M. Wetklo. 2008c. Population structure and stock identification of chum salmon Oncorhynchus keta from British Columbia determined with microsatellite DNA variation. Can. J. Zool. 86:1002-1014. Beacham, T. D., B. McIntosh, C. MacConnachie, K. M. Miller, R. E. Withler, and N. V. Varnavskaya. 2006a. Pacific Rim population structure of sockeye salmon as determined from microsatellite analysis. Trans. Am. Fish. Soc. 135:174-187. Beacham, K. L. Jonsen, J. Supernault, M. Wetklo, L. Deng, and N. Varnavskaya. 2006b. Pacific Rim population structure of Chinook salmon as determined from microsatellite variation. Trans. Am. Fish. Soc. 135:1604-1621. Beacham, T. D., N. V. Varnavskaya, K. D. Le, and M. Wetklo. 2008a. Determination of population structure and stock identification of chum salmon (Oncorhynchus keta) in Beacham et al.: Population structure of Oncorhynchus keta across the Pacific Rim 259 Russia determined with microsatellites. Fish. Bull. 106:245-256. Brykov, V. A., N. E. Poliakova, and A. V. Prokhorova. 2003. Phylogenic and geographic analysis of chum salmon Oncorhynchus keta (Walbaum) in Asian populations based on mitochondrial DNA variation. Russ. J. Genet. 39:61-67. Buchholz, W. G., S. J. Miller, and W. J. Spearman. 2001. Isolation and characterization of chum salmon microsatellite loci and use across species. Anim. Genet. 32:160-165. Cairney, M., J. B. Taggart, and B. Hoyheim. 2000. Characterization of microsatellite and minisatel- lite loci in Atlantic salmon ( Salmo salar L.) and cross- species amplification in other salmonids. Mol. Ecol. 9:2175-2178. Cavalli-Sforza, L. L., and A. W. F. Edwards. 1967. Phylogenetic analysis: models and estimation procedures. Am. J. Hum. Genet. 19:233-257. Chen, J.-P., D.-J. Sun, C.-Z. Dong, B. Liang, W.-H.Wu, and S.-Y. Zhang. 2005. Genetic analysis of four wild chum salmon Oncorhyn- chus keta populations in China based on microsatellite markers. Environ. Biol. Fishes 73:181-188. Efremov, V. V. 2001. Genetic variation and differentiation of popula- tions of chum salmon Oncorhynchus keta (Walbaum) from Southern Russian Far East. Russ. J. Genet. 37:283-289. Estoup, A., P. Jarne, and J. M. Cornuet. 2002. Homoplasy and mutation model at microsatel- lite loci and their consequences for population genetics analysis. Mol. Ecol. 11:1591-1604. Felsenstein J. 1993. PHYLIP: Phylogeny Inference Package. Univ. Washington, Seattle, WA. Ginatulina, L. K. 1992. Genetic differentiation among chum salmon, Oncorhynchus keta (Walbaum), from Primorye and Sakhalin. J. Fish Biol. 40:33-38. Goudet, J. 1995. FSTAT (version 1.2): A program for IBM PC com- patibles to calculate Weir and Cockerham’s (1984) esti- mators of F-statistics. J. Hered. 86:485-486. Hopkins, D. M. 1967. The Cenozoic history of Beringia-a synthesis. In The Bering land bridge. (D. M. Hopkins, ed.) p. 451- 484. Stanford Univ. Press, Stanford, CA. Kaeriyama M. 1999. Hatchery programmes and stock management of salmonid populations in Japan. In Stock enhancement and sea ranching (B. R. Howell, E. Moksness, and T. Svasand, eds.). p.153-167. Blackwell Science, Oxford, U.K. Kalinowski, S. T. 2005. Do polymorphic loci require large sample sizes to estimate genetic distances? Heredity 94:33-36. Kijima A, and Y. Fujio. 1982. Correlation between geographic distance and genetic distance in populations. Bull. Jap. Soc. Sci. Fish. 48:1703-1709. Knebel, H. J., and J. S. Creager. 1973. Yukon River: evidence for extensive migration during the Holocene transgression. Science 179:1230- 1232. Kondzela, C. M., C. M. Guthrie, S. L. Hawkins, C. D. Russell, J. H. Helle, and A. J. Gharrett. 1994. Genetic relationships among chum salmon pop- ulations in southeast Alaska and northern British Columbia. Can. J. Fish. Aquat. Sci. (suppl. 1):50- 64. Lindsey, C. C., and J. D. McPhail. 1986. Zoogeograpy of fishes of the Yukon and Macken- zie basins. In The zoogeography of North American freshwater fishes (C. H. Hocutt and E. O. Wiley, eds.) p. 639-674. Wiley, New York. McPhail, J. D., and C. C. Lindsey. 1970. Freshwater fishes of northwestern Canada and Alaska. Bull. Fish. Res. Board Can. 173:1-381. Nelson, R. J., and T. D. Beacham. 1999. Isolation and cross species amplification of micro- satellite loci useful for study of Pacific salmon. Anim. Genet. 30:228-229. Okazaki, T. 1982a. Geographical distribution of allelic variation of enzymes in chum salmon Oncorhynchus keta , river popu- lations of Japan and the effects of transplantation. Bull. Jap. Soc. Sci. Fish. 48:1525-1535. Okazaki, T. 1982b. Genetic study on population structure in chum salmon (Oncorhynchus keta). Bull. Far Seas Fish. Res. Lab. 19:25-116. Olsen, J. B., S. L. Wilson, E. J. Kretschmer, K. C. Jones, and J. E. Seeb. 2000. Characterization of 14 tetranucleotide microsat- ellite loci derived from sockeye salmon. Mol. Ecol. 9:2185-2187. Park, L. K, M. A. Brainard, D. A. Dightman, and G. A. Winans. 1993. Low levels of intraspecific variation in the mitochon- drial DNA of chum salmon (Oncorhynchus keta). Mol. Mar. Biol. Biotech. 2:362-370. Perriere, G., and M. Gouy. 1996. WWW-query: An on-line retrieval system for bio- logical sequence banks. Biochimie 78:364-369. Phelps, S. R., L. L. LeClair, S. Young, and H. L. Blankenship. 1994. Genetic diversity patterns of chum salmon in the Pacific Northwest. Can. J. Fish. Aquat. Sci. (suppl. 1 ) : 65 — 83 . Polyakova, N. E., A. V. Semina, and V. A. Brykov. 2006. The variability in chum salmon Oncorhynchus keta (Walbaum) mitochondrial DNA and its connec- tion with the paleogeological events in the northwest Pacific. Russ. J. Genet. 42:1164-1171 Rexroad, C. E., R. L. Coleman, A. M. Martin, W. K. Hershberger, and J. Killefer. 2001. Thirty-five polymorphic microsatellite markers for rainbow trout (Oncorhynchus mykiss). Anim. Genet. 32:317-319. Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution 43:223-225. Salo, E. O. 1991. Life history of chum salmon (Oncorhynchus keta). In Life history of Pacific salmon (C. Groot and L. Margolis, eds.), p. 231-309. UBC Press, Vancou- ver, BC. Salmenkova, E. A., V. T. Omelchenko, D. V. Politov, K. I. Afa- nasyev, and G. A. Rubtsova. 2007. Genetic diversity of northern Okhotsk Sea coast populations of chum salmon with natural and artificial reproduction. Biologiya Morya 33:299-308. 260 Fishery Bulletin 107(2) Sato, S., J. Ando, H. Ando, S. Urawa, A. Urano, and S. Abe. 2001. Genetic variation among Japanese populations of chum salmon inferred from the nucleotide sequences of the mitochondrial DNA control region. Zool. Sci. (Tokyo) 18: 99-106. Sato, S., H. Kojima, J. Ando, H. Ando, R. L. Wilmot, L. W. Seeb, V. Efremov, L. LeClair, W. Buchholz, D.-H. Jin, S. Urawa, M. Kaeriyama, A. Urano, and S. Abe. 2004. Genetic population structure of chum salmon in the Pacific Rim inferred from mitochondrial DNA sequence variation. Environ. Biol. Fishes 69:37-50. Seeb, L. W., and P. A. Crane. 1999. High genetic heterogeneity in chum salmon in Western Alaska, the contact zone between northern and southern lineages. Trans. Am. Fish. Soc. 128:58-87. Small, M. P., T. D. Beacham, R. E. Withler, and R. J. Nelson. 1998. Discriminating coho salmon ( Oncorhynchus kisutch) populations within the Fraser River, British Colum- bia using microsatellite DNA markers. Mol. Ecol. 7:141-155. Smith, C. T., B. F. Koop, and R. J. Nelson. 1998. Isolation and characterization of coho salmon (Oncorhynchus kisutch) microsatellites and their use in other salmonids. Mol. Ecol. 7:1613-1621. Smith, C. T., and L. W. Seeb. 2008. Number of alleles as a predictor of the relative assignment accuracy of short tandem repeat (STR) and single-nucleotide-polymorphism (SNP) baselines for chum salmon. Trans. Am. Fish. Soc. 137:751-762. Spies, I. B., D. J. Brasier, P. T. L. O’Reilly, T. R. Seamons, and P. Bentzen. 2005. Development and characterization of novel tetra-, tri-, and dinucleotide microsatellite markers in rain- bow trout (Oncorhynchus mykiss). Mol. Ecol. Notes 5:278-281. Taylor, E. B., T. D. Beacham, and M. Kaeriyama. 1994. Population structure and identification of North Pacific Ocean chum salmon ( Oncorhynchus keta) revealed by an analysis of minisatellite DNA variation. Can. J. Fish. Aquat. Sci. 51:1430-1442. Varnavskaya, N. V., C. C. Wood, and R. J. Everett. 1994a. Genetic variation in sockeye salmon (Oncorhynchus nerka ) populations of Asia and North America. Can. J. Fish. Aquat. Sci. 51(suppl. 1):132-146. Waples, R. S. 1990. Temporal changes of allele frequency in Pacific salmon populations: implications for mixed-stock fishery analysis. Can. J. Fish. Aquat. Sci. 47:968-976. Warner, B. C., R. W. Mathews, and J. J. Clague. 1982. Ice-free conditions on the Queen Charlotte Islands, British Columbia, at the height of the late Wisconsin glaciation. Science 218:675-677. Weir, B.S., and C. C. Cockerham. 1984. Estimating F-statistics for the analysis of popula- tion structure. Evolution 38:1358-1370. Williamson, K. S., J. F. Cordes, and B. P. May. 2002. Characterization of microsatellite loci in chinook salmon (Oncorhynchus tshawytscha ) and cross-species amplification in other salmonids. Mol. Ecol. Notes 2:17-19. Wilmot, R. L., R. J. Everett, W. J. Spearman, R. Baccus, N. V. Varnavskaya, and S. V. Putivkin. 1994. Genetic stock structure of Western Alaska chum salmon and a comparison with Russian Far East stocks. Can. J. Fish. Aquat. Sci. 51 (suppl. l):84-94. Winans, G. A., P. B. Aebersold, S. Urawa, and N. V. Varnavskaya. 1994. Determining continent of origin of chum salmon ( Oncorhynchus keta ) using genetic identification tech- niques: status of allozyme baseline in Asia. Can. J. Fish. Aquat. Sci. 51(suppl. 1):95-113. Wood, C. C. 1995. Life history variation and population structure in sockeye salmon. Am. Fish. Soc. Symp. 17:195-216. 261 Corrigendum Fishery Bulletin 105(3), p. 404. Stark, James W. Geographic and seasonal variations in maturation and growth of female Pacific cod ( Gadus macrocepho/us) in the Gulf of Alaska and Bering Sea Page 404. The x-axis scale in Figure 5 should read as follows: 262 Fishery Bulletin 107(2) Fishery Bulletin Guidelines for authors Manuscript Preparation Contributions published in Fishery Bulletin describe original research in marine fishery science, fishery engineering and economics, as well as the areas of marine environmental and ecological sciences (including modeling). Preference will be given to manuscripts that examine processes and underlying patterns. Descriptive reports, surveys, and observational papers may occa- sionally be published but should appeal to an audience outside the locale in which the study was conducted. Although all contributions are subject to peer review, responsibility for the contents of papers rests upon the authors and not on the editor or publisher. Submission of an article implies that the article is original and is not being considered for publication elsewhere. Articles may range from relatively short contributions (10-15 typed, double-spaced pages, tables and figures not included) to extensive contributions (20-30 typed pages). Manu- scripts must be written in English; authors whose native language is not English are strongly advised to have their manuscripts checked by English-speaking col- leagues before submission. Title page should include authors’ full names and mailing addresses and the senior author’s telephone, fax number, and e-mail address, and a list of key words to describe the contents of the manuscript. Abstract should be limited to 200 words (one-half typed page), state the main scope of the research, and emphasize the author’s conclusions and relevant findings. Do not review the methods of the study or list the contents of the paper. Because abstracts are circulated by abstract- ing agencies, it is important that they represent the research clearly and concisely. Text must be typed in 12 point Times New Roman font throughout. A brief introduction should convey the broad significance of the paper; the remainder of the paper should be divided into the following sections: Materials and methods, Results, Discussion (or Conclusions), and Acknowl- edgments. Headings within each section must be short, reflect a logical sequence, and follow the rules of multi- ple subdivision (i.e., there can be no subdivision without at least two items). The entire text should be intelligible to interdisciplinary readers; therefore, all acronyms, abbreviations, and technical terms should be written out in full the first time they are mentioned. Include FAO common names for species in the list of keywords and in the introduction. Regional common names may be used throughout the rest of the text if they are dif- ferent from FAO common names which can be found at http://www.fishbase.org/search.html. Follow the U.S. Government Printing Office Style Manual (1984 ed.) and the CBE Style Manual (6th ed.) for editorial style; for fish nomenclature follow the most current issue of the American Fisheries Society’s Common and Scientific Names of Fishes from the United States and Canada. Dates should be written as follows: 11 November 2000. Measurements should be expressed in metric units, e.g., 58 metric tons (t); if other units of measurement are used, please make this fact explicit to the reader. Write out the numbers zero through nine unless they form part of measurement units (e.g., nine fish but 9 mm). Refrain from using the shorthand slash (/), an ambiguous symbol, in the general text Literature cited comprises published works and those accepted for publication in peer-reviewed literature (in press). Follow the name and year system for citation format in the “Literature cited” section (that is say, citations should be listed alphabetically by the authors’ last names, and then by year if there is more than one citation with the same authorship). If there is a sequence of citations in the text, list chronologically: (Smith, 1932; Green, 1947; Smith and Jones, 1985). Abbrevia- tions of serials should conform to abbreviations given in the Serial Sources for the BIOSIS Previews Database. Authors are responsible for the accuracy and complete- ness of all citations. Literature citation format: Author (last name, followed by first-name initials). Year. Title of report or manuscript. Abbreviated title of the series to which it belongs. Always include number of pages. Cite all software and special equipment or chemical solutions used in the study, not in a footnote but within parentheses in the text (e.g., SAS, vers. 6.03, SAS Inst., Inc., Cary, NC). Tables are often overused in scientific papers; it is seldom necessary or even desirable to present all the data associated with a study. Tables should not be excessive in size and must be cited in numerical order in the text. Headings should be short but ample enough to allow the table to be intelligible on its own. All unusual symbols must be explained in the table legend. Other incidental comments may be footnoted with italic numeral footnote markers. Use asterisks to indicate probability in statistical data. Do not type table legends on a separate page; place them above the table data. Do not submit tables in photo mode. • Zeros should precede all decimal points for values less than one. • Sample size, n, should be italicized. • Capitalize the first letter of the first word in all labels within figures. • Do not use overly large font sizes in maps and for units of measurements along axes in figures. • Do not use bold fonts or bold lines in figures. • Do not place outline rules around graphs. • Do not use horizontal lines through graphs to indicate measurement units. • Use a comma in numbers of five digits or more (e.g. 13,000 but 3000). • Maps require a North arrow and degrees latitude- longitude (e.g., 170°E). 263 Figures include line illustrations, photographs (or slides), and computer-generated graphs and must be cited in numerical order in the text. Graphics should aid in the comprehension of the text, but they should be limited to presenting patterns rather than raw data. Figures should not exceed one figure for every four pages of text. Figures must be labeled with author’s name and number of the figure. Avoid placing labels vertically (except of y axis). Figure legends should explain all symbols and abbreviations and should be double-spaced on a separate page at the end of the manuscript. Please note that we do not print graphs in color. Failure to follow these guidelines and failure to correspond with editors in a timely manner will delay publication of a manuscript. Copyright law does not apply to Fishery Bulletin , which falls within the public domain. However, if an author reproduces any part of an article from Fishery Bulletin in his or her work, reference to source is consid- ered correct form (e.g., Source: Fish. Bull 97:105). Submission The Scientific Editorial Office encourages authors to submit their manuscripts as a single PDF (preferred) or Word (zipped) document by e-mail to Fishery. Bulletin@noaa.gov. Please use the subject heading, “Fishery Bulletin manuscript submission”. Do not send encrypted files. For further details on electronic sub- mission, please contact the Scientific Editorial Office directly (see address below). Or you may send your manuscript on a compact disc in one of the above formats along with four printed copies (one original plus three copies [stapled]) to the Scientific Editor, at the address shown below. Richard D. Brodeur, Ph.D. Scientific Editor, Fishery Bulletin Northwest Fisheries Science Center 2030 S. Marine Science Dr. Newport, Oregon 97365-5296 Once the manuscript has been accepted for publication, you will be asked to submit a final electronic copy of your manuscript. When requested, the text and tables should be submitted in Word or Word Rich Text Format. Figures should be sent as PDF files, Windows metafiles, tiff files, or EPS files. Send a copy of figures in the origi- nal software if conversion to any of these formats yields a degraded version. Questions? If you have questions regarding these guidelines, please contact the Managing Editor, Sharyn Matriotti, at Sharyn.Matriotti@noaa.gov Questions regarding manuscripts under review should be addressed to Richard Brodeur, Scientific Editor, at Rick.Brodeur@noaa.gov. Fishery Bulletin Subscription form Superintendent of Documents Publications Order Form *5178 I I YES, please send me the following publications: Subscriptions to Fishery Bulletin for $36.00 per year ($50.40 foreign) The total cost of my order is $ . Prices include regular domestic postage and handling and are subject to change. (Company or Personal Name) (Please type or print) (Additional address/attention line) (Street address) (City, State, ZIP Code) (Daytime phone including area code) (Purchase Order No.) Charge your order. ITS EASY! Please Choose Method of Payment: | Check Payable to the Superintendent of Documents ] GPO Deposit Account ] VISA or MasterCard Account To fax your orders (202) 512-2250 (Credit card expiration date) (Authorizing Signature) Mail To: Superintendent of Documents P.O. Box 371954, Pittsburgh, PA 15250-7954 Thank you for your order! Also available online at http://bookstore.gpo.gov/collections/fishery-bulletin SET>S \$m ?<§& SMITHSONIAN INSTITUTION LIBRARIES 3 9088 0 459 5797