OCEANOGRAPHY and MARINE BIOLOGY: AN ANNUAL REVIEW (Volume 46) - Chapter 8 (end) doc

These hypotheses dramatically influenced the direction of research in the field by 1 focusing research into the causes of interannual variability in recruitment on the egg and larval sta

WILLIAM C LEGGETT1 & KENNETH T FRANK2

1 Department of Biology, Queen’s University, Kingston, ON K7L 3N6, Canada

E-mail: wleggett@queensu.ca

2 Department of Fisheries and Oceans, Bedford Institute

of Oceanography, Dartmouth NS B2Y 4A2, Canada

E-mail: FrankK@mar.dfo-mpo.gc.ca

Abstract The development of the field of fisheries oceanography over the past century has

been heavily influenced by a relatively small number of paradigms that have shaped thinking, influenced lines of enquiry and occasionally stalled progress in the field This review provides

an overview of what are considered to be the most influential paradigms in the discipline Each begins with a brief discussion of its origins Next their respective (and often overlapping) impact

on the development of the discipline is discussed and then the evolution of these paradigms

as shaped by new advances in approaches and technologies and by direct challenges to their underlying assumptions is reviewed For each, the endpoint is an overview of the current state

of knowledge and thinking and the probable future direction of research in the area The review concludes with an overview of the probable future directions of research in the discipline as a whole

Introduction

The discipline of modern fisheries oceanography has its origins in the work of Johan Hjort (1914, 1928), who was the first to formally hypothesize a link between the dynamics of fish populations and the dynamics of their environment In his ‘critical period’ paradigm Hjort argued that variabil-ity in food availability during the transition from endogenous to exogenous feeding in larval fishes, typically a very narrow time window, was central to the survival of individual larval cohorts Hjort hypothesized that when food was abundant, survival (and recruitment) would be high, and when food was scarce survival and recruitment would be low This hypothesis was subsequently general-ized by Cushing (1975) who reasoned, in his ‘match mismatch’ hypothesis, that food availability was linked to the interaction between interannual differences in the timing of spawning/hatching and the timing and magnitude of the primary and secondary production cycles in the ocean These hypotheses dramatically influenced the direction of research in the field by (1) focusing research into the causes of interannual variability in recruitment on the egg and larval stages of fishes, (2) provid-ing a simplifying construct within which to explore the causes of temporal changes in the abun-dance of commercially important marine fishes and (3) by linking fluctuations in the abundance of fishes directly to the dynamics of other components of the ocean ecosystem In hindsight, Hjort’s major contribution appears to have been to awaken thinking and research into the nature of this dynamic interaction between fish and their environment Prolonged adherence to Hjort’s ideas, and the lure of Cushing’s hypothesis, combined to dominate thinking and research in the field for most

of the twentieth century, some would say negatively (Leggett & DeBlois 1994) As the discipline has advanced several new paradigms have evolved While originally offered as simplifying constructs,

and as the foundation for new studies of important processes governing the dynamics of fish stocks, some of these new paradigms now approach the status of dogma once accorded Hjort’s hypothesis and have the potential, if pursued uncritically, to once again delay progress in the field This review examines the major paradigms that have shaped thinking and research in fisheries oceanography over the past century and explore how recent research has affected understanding of their validity and usefulness now, and going forward Table 1 provides an overview of these paradigms ordered hierarchically from individuals to ecosystems.

The need for a more holistic and effective approach to the science of fisheries oceanography, and of the management strategies applied to marine fishes, is evident from the breadth of species now undergoing serious decline, not only worldwide, but also within more restricted ecosystems (Dower et al 2000, Figure 1)

The advent of new methods (ageing techniques for larval fishes, the application of biochemical and molecular techniques, the development of instrumentation and computing capacity that has allowed physical oceanographers to measure and model the highly dynamic and variable environ-ment of continental shelf ecosystems, satellite imagery, etc.) have created new opportunities to explore the dynamics of marine fishes in relation to the dynamics of their physical environment and the ecosystems they inhabit

Paradigm 1: Spawning stock biomass is a suitable proxy for the reproductive potential of a stock

The Hjort and Cushing models focused primarily on the factors determining interannual variability

in mortality rates experienced by larvae (the life stage widely believed to be the most vulnerable to environmental variability (reviewed in Leggett & DeBlois 1994) Another highly influential para-digm holds that it is the product of the number of offspring generated by a spawning stock and the rate of mortality experienced by those offspring until they recruit to the reproducing population that govern the dynamics of fish populations This thinking is inherent in stock recruitment models (Ricker 1954, Beverton & Holt 1957) that have formed the cornerstone of studies of the recruitment dynamics of fishes, and of management based on these studies, for most of the latter half of the twentieth century These models assume strong density dependence (Figure 2)

Fundamental to the practical application of these models has been the assumption that the total spawning stock biomass (SSB) is an acceptable proxy for the reproductive potential of the stock of interest This assumption is employed because SSB can be readily derived from fisheries survey data, whereas total egg production (TEP), a more reliable and realistic indicator of reproductive output is more labour intensive to obtain and, for this reason, does not exist for most stocks.The use of SSB as a proxy for TEP implies that (1) spawner biomass is proportional to the TEP and (2) in lay terms, “an egg is an egg is an egg,” that is, all eggs are equal in their potential to

Table 1 Paradigms in fisheries oceanography

1 Spawning stock biomass (SSB) is a suitable proxy for the reproductive potential of a stock.

2 Marine fish eggs and larvae are generally designed for dispersion and potential colonization (panmixia).

3 In marine temperate systems, fish spawn in springtime so that peak larval abundance coincides with maximum prey availability (Cushing’s match/mismatch hypothesis).

4 Environmentally based recruitment models, when updated with new data, invariably fail; recruitment prediction is an intractable problem, particularly when it is based on processes associated with the growth and mortality of the early life-history stages.

5 Populations cannot irreversibly collapse/collapsed populations will recover in the absence of fishing.

6 Fish stocks can be managed in isolation from their total environment/habitat.

7 Population recovery is synonymous with rebuilding.

produce a recruit to the population However, the variation in SSB is generally insufficient to account for the related variation observed in recruitment in marine fish populations (Shepherd et al 1984, Wooster & Bailey 1989, Rijnsdorp et al 1991, Marshall et al 1998) Notwithstanding this reality, the paradigm that SSB is a suitable proxy for total reproductive effort has persisted and has, in turn, been an important contributor to the persistence of the Hjort and Cushing hypotheses and the focus

of research into the causes of recruitment variation in marine fishes This includes the search for ecological and/or environmental factors affecting egg and larval mortality — an approach that has proved challenging and, in only limited cases, successful (see Paradigm 4) This has impeded the development of a coherent understanding of the recruitment dynamics of marine fishes in particular and the science of fisheries management in general (Rothschild 1986, Hilborn & Walters 1992)

cod 3NO spring-SSB

cod 2J3KL-Offhore Survey-SSB

white hake 4T-SSB

thorny skate 4T-SSB

cod 4Vn-SSB

cod 4X-SSB cod 4T-SSB smooth skate 4T-SSB

haddock 4VW-SSB

mackerel-SSB winter skate 4T-SSB

haddock 4X-SSB

plaice 3Ps-SSB haddock 5Z-SSB

thorny skate 4VWX-SSB

cod 4VsW-SSB pollack 4VWX-SSB

Figure 1 (See also Colour Figure 1 in the insert following p 250.) ordination of the time of series of ing stock biomass of various species from scientific surveys conducted throughout the north-west Atlantic, illustrating that the majority of the stocks are at biomass levels well below (red) the long-term average Intensity of colours is proportional to the magnitude of the standardized anomaly in standard deviation units Alphanumeric labelling refers to the Northwest Atlantic Fisheries organization management unit.

spawn-In the 1990s these assumptions began to come under increasing scrutiny (Lowerre-Barbieri

et al 1998, Marshall et al 1998, Marteinsdottir & Steinarsson 1998, Marshall & Frank 1999,

Marshall et al 1999) Marshall et al (1998) were the first to demonstrate in situ the potential fallacy

of the assumption inherent in the use of SSB as a surrogate for TEP Their work with Barents Sea

cod (Gadus morhua) showed that recruitment was positively correlated with the quantity of lipid

energy stored in the liver of mature females or, in other words, with female condition Boyd et al (1998) observed a similar relationship between parental fat levels and recruitment in Cape anchovy

(Engraulis capensis) In cod, total liver energy was proportional to TEP and, like TEP, varied with capelin (Mallotus villosus) abundance (yaragina & Marshall 2000) such that by either measure the

reproductive potential of a fixed number of mature females was significantly higher when the tity of food available to maturing fish was abundant and correspondingly lower when it was scarce

quan-In contrast, cod SSB was found to be not statistically different at high and low prey levels quan-In short, their study confirmed prior suspicions regarding the paradigm that SSB is an inadequate surrogate for TEP and demonstrated that replacing SSB with more accurate measures of reproductive poten-tial is fundamental to a fuller understanding of the dynamics of recruitment in marine fishes.The work of Marshall et al (1998, 1999) has been an important factor in redirecting the focus

of studies into the causes of interannual variation of recruitment in marine fish populations Their findings demonstrated that a more successful pursuit of process-oriented models of the recruitment dynamics of marine fish will require integration of the processes affecting reproductive output (growth, production and condition of spawners) with mortality processes affecting egg, larval and juvenile survival (Ulltang 1996, Marshall et al 2000)

Research by yaragina & Marshall (2000) has demonstrated that the search for a comprehensive understanding of the factors regulating variability in reproductive potential, independent of SSB, may be as complex, and as inextricably tied to environmental and ecological factors, as are the causes of variation in egg and larval survival For example, their study of temporal variation in the liver condition index (LCI) — an important determinant of reproductive potential (Marshall et al

1999) — of five length classes of north-east Arctic cod (Gadus morhua) showed that while

varia-tions in the abundance of capelin, a major food source, was the proximate determinant of variation

in LCI, an indirect, but perhaps not ultimate, determinant of variation in the index appears to have

been the abundance of herring (Clupea harengus), which influence cod LCI indirectly via their

predation on capelin, which in turn are the main prey of cod

Interest in the role and importance of maternal effects as regulators of the recruitment dynamics

of marine fishes has increased since the publication of these findings Scott et al (2006) modelled the daily reproductive output of a range of simulated age/size-structured populations of Atlantic

North Sea herring

Figure 2 Ricker stock recruitment model fit to data for North Sea herring SSB, spawning stock biomass.

cod, created under contrasting stock recruitment scenarios, over an entire spawning season The objective was to determine the effects of individual female condition and egg quality on stock repro-ductive potential (SRP) Their findings suggest that, in two populations having equal SSB, the effect

of low levels of individual condition in one can lead to almost total reproductive failure for that population Moreover, the positive effect of increased individual condition was found to depend on

the particular population structure designated in the model, a variable that, in situ, can be strongly

influenced by both fishing intensity and environment Indeed, they reported that differences in population size and age structure produced by differences in fishing mortalities ranging from 0 to 1.0 could reduce the SRP by 48–74% depending on the related assumptions regarding female con-dition and quality Any factor, anthropogenic or environmental, inducing changes in mortality of equivalent magnitude would be expected to produce a similarly scaled result

Vallin & Nissling (2000) studied the effect of female age on the survival of cod eggs and vae in the Baltic, where successful spawning is restricted to the deep basins at salinities varying between 11 and 20 ppm Due to the oxygen depletion commonly prevailing in these areas, neutral egg buoyancies above oxygen-critical depths are important to egg survival They found that large females produce larger eggs that exhibit neutral egg buoyancy at lower salinities, thereby ensuring egg development in more favourable oxygen conditions The number of recruited cod (age 2), in the Baltic Sea over two eras characterized by different conditions (1967–1980 and 1981–1994), was found to be positively related to the fraction of eggs produced by older females (5+ yr), implying a strong maternal effect on recruitment Berkeley et al (2004) report a similar positive relationship

lar-between female age and probability of egg survival in longnose grenadier (Coelorhynchus tus) Marteinsdottir & Steinarsson (1998) report similar findings for the egg and early larval stages

carmina-of cod

The findings of Vallin & Nissling (2000) also illustrate the potential for negative effects of fishery-induced changes in population age structure on population size and persistence Scott et al (1999) modelled this fishing effect and found that the effects of the loss of more fecund older/larger individuals in the population could lead to overestimation of the number of potential recruits to populations experiencing higher levels of fishing mortality by as much as 60%

Environmental variables can also influence the condition of females and the demographics of populations in ways that can dramatically alter reproductive output independent of SSB Scott et al (1999) note that when size- (growth-) related maternal effects on egg viability were incorporated into their fishing effect model, the number of potential recruits to heavily exploited populations could be reduced by a further 10%

The work of Choi et al (2004) illustrates just how profound these environmental influences can be In their paper on the devolution of the Scotian Shelf ecosystem off Nova Scotia, Canada, they document declines in growth rates, size at age and age at maturation of the entire benthic fish community that mirror the changes modelled and documented above These changes occurred pro-gressively over a 45-yr time period (1960–2005), the most dramatic changes occurring in the early

1990s They resulted in average sizes of mature (age 5) cod, haddock (Melanogrammus aeglefinus), pollock (Pollachius virens) and silver hake (Merluccius bilinearis) that were 70–80% of the 1970s

levels and remain depressed in spite of a moratorium on fishing that began in 1993 (Figure 3) Drinkwater (2002) estimates that approximately 30–50% of the decline in SSB of cod that occurred

in the 1980s and early 1990s was linked to these reductions in deep-water temperatures and their depressing effect on growth rates and corresponding sizes at age Given the strong positive relation-ship between body size and fecundity in these species (Pinhorn 1984, Waiwood & Buzeta 1989), a

corresponding decline in per individual reproductive potential, independent of losses due to

coinci-dent reductions in the total number of spawners, clearly occurred

Their analysis also revealed that these dramatic changes in reproductive output were not simply the consequence of reductions in the size of reproducing adults (a product of the negative effects that quotas based on biomass and the behaviour of fishermen focusing their efforts on the largest fish available; e.g., see Drinkwater 2002) For example, during the 1960s and 1970s the physiologi-cal condition of these fishes exceeded the 45-yr norm However, throughout the 1980s and 1990s physiological condition declined sharply and by 2000 the majority of the fish were in below-average condition Given the documented impact of female condition on egg quality and recruitment in cod (Marshall et al 1999), and the evidence of maternal effects on survival linked to female size and condition, it is likely that in this species, and perhaps others, a further reduction in effective reproductive output, independent of SSB, was experienced as a consequence of combined effects of declining size and condition, both of which had an environmental component.

There is also evidence that the total metabolic rate of the entire demersal ecosystem was depressed as a consequence of changes in ocean climate These ocean climate changes included

a progressive decline in bottom temperatures, an increase in the volume of the cold intermediate layer, and movement of the Gulf Stream Front to a more offshore position These changes are judged

to have been caused by an increased along-shelf advection from the Gulf of St Lawrence and ern Newfoundland augmented by local, atmospherically induced cooling (Drinkwater et al 2003) one of the outcomes of this change was a dramatic reduction in the biomass of the once-dominant

south-euphausiid (Euphausia superba) and a corresponding decline in their contribution to the overall diet

of the demersal fish assemblage For example, the contribution of this euphausid species to the diet

of pollock and cod declined from >65% by weight in the 1980s to <10% in the 1990s (Hanson & Chouinard 2002, Carruthers et al 2005)

The resulting hysteresis in the structure of the Scotian Shelf ecosystem in the late 1980s duced an ecosystem dominated by smaller pelagic fishes and benthic invertebrates (mainly shrimp and crab) in which the once-dominant large-bodied demersal fishes now play a relatively minor role The fact that these once-dominant species have not recovered even in the absence of exploitation,

pro-20 30 40 50 60 70

1970 1975 1980 1985 1990 1995 2000

Cod Haddock Pollock Silver hake

Figure 3 (See also Colour Figure 3 in the insert.) Changes with time in lengths at age 5 for four groundfish species from the eastern Scotian Shelf.

as traditional stock recruitment models would predict, suggests that a fundamental change in their fitness has occurred The disjunction between SSB and effective reproductive output appears to be a significant factor in this change in fitness and may have contributed in a major way to the dramatic collapse of these once-dominant demersal species in the early 1990s.

Paradigm 2: Marine fish eggs and larvae are generally

designed for dispersion and potential colonization (panmixia)

Reproduction in commercially exploited marine species generally involves the spawning of billions

of tiny eggs that float in the near-surface waters during the spring of each year The generality and magnitude of these events were widely considered to be an adaptation to the very high mortality of early life stages caused by the vagaries of the environment (Rothschild 1986, Rothschild & DiNardo 1987) one variable identified as important in this context was ocean currents, often driven by wind, which caused larvae to be transported to regions where food and other factors were less favourable Several correlation studies reported significant relationships between larval survival, measured by the abundance of the recruiting year-class, and dispersive events

For example, it has been hypothesized that warm core rings (eddies representing instabilities in the Gulf Stream that can entrain large volumes of water from the continental shelf) can transport sufficient numbers of eggs and larvae off the shelf to have a significant negative impact on recruit-ment Myers & Drinkwater (1989), who examined this hypothesis using estimates of entrainment from 14 yr of satellite imagery and recruitment estimates for 25 fish and shellfish stocks ranging from the Mid-Atlantic Bight to the southern Grand Banks, found that increased warm core ring activity was associated with low recruitment levels in 17 of 18 groundfish stocks examined The sole exception was cod on Georges Bank While the level of significance related to each stock was low, the collective result was consistent with the original hypothesis For other examples of the possible link between dispersive effects and recruitment see Carruthers et al (1951) and Bailey (1981) The findings of these and other studies led to the view that either excessive dispersion of larvae or dis-placement to areas distant from the nursery ground was a primary cause of the reduced recruitment that resulted when wind strengths and direction were unfavourable However, in most cases updated analyses of these models have failed to support the relationships originally observed (Myers 1998, see Paradigm 4)

The seminal work of Sinclair (1988) showed that many continental shelf spawning locations of marine fishes were located in areas characterized by retention features or semipermanent gyres, the current characteristics of which act to minimize the advection of the egg and larval stages and con-sequently their vulnerability to dispersive processes Many subsequent studies have expanded the suite of species that utilize relatively non-dispersive oceanographic settings for spawning and early larval development (Hinrichsen et al 2001, North & Houde 2001, Lett et al 2007) In addition, the advent of technologies allowing high-resolution, discrete-depth sampling of the water column revealed that ontogenetic shifts in depth distributions and vertical migratory behaviour of larvae also serve to minimize dispersion from the spawning areas These findings have moderated the belief that the early life stages of marine species were simply passive drifters and that survival was largely dependent on the “mercy” of the currents

one of the most striking examples of this non-dispersive reality involves haddock that spawn

on the offshore banks of the Scotian Shelf Because of the recirculation that occurs around each

of these discrete spawning banks there is a strong tendency for the egg and larval distributions of haddock that spawn there to be discrete And, when larvae metamorphose to the juvenile stage, a development process that takes about 90 days, settlement to the bottom occurs in the same zones in which they were spawned, which now become prime feeding areas for the juvenile and adult stages (Frank et al 2000)

These findings are consistent with the results of modelling studies of the scales of connectivity

in Caribbean reef fishes (Cowen et al 2006) They found that larval dispersal scales for most species were of the order of 50–100 km, much smaller than has generally been assumed Their analyses also indicate that passive dispersal/retention is insufficient for population replenishment, recruitment levels generated from the use of passive dispersal assumptions in the model being one to two orders

of magnitude lower than that required for successful population replenishment

These studies call for a general rethinking of the adaptive character of the spawning behaviour

of many marine species from spawning as a vehicle for colonization or bet hedging to one in which the entire early life-history period is viewed as an elegant suite of adaptations to an environment that

is highly variable at interannual and smaller timescales but is more predictable in the longer term.However, while this longer-term predictability is reflected in adaptations involving both the timing and location of spawning, interannual variability in the ocean environment can result in important disruptions to their efficacy that lead to recruitment variability and even recruitment fail-ure Well-known examples include the relationship between capelin recruitment and onshore wind discovered by Leggett et al (1984), the destabilizing effects of warm core rings on the retention dynamics of Scotian Shelf banks (Myers & Drinkwater 1989) and the influence of upwelling events

on recruitment processes along the North African coast (Cury & Roy 1989) Such disruptive events can also lead to the establishment of new subpopulations through colonization of new habitats when environmental conditions permit Detection of such colonization episodes is most evident when they occur outside the normal range of distribution The larval drift associated with anomalous oceano-graphic conditions involving the displacement of cod larvae from Iceland to West Greenland — a colonization event covering a distance of almost 1000 km (Frank 1992) — is a striking case in point

A similar displacement, on a much smaller geographic scale, was shown for haddock larvae ing from the eastern Scotian Shelf Larvae of this species episodically drift downstream to Browns Bank on the western Scotian Shelf In the case of both cod and haddock, the dispersed larval stages appear to persist and to contribute to the recruiting year-class in the colonized area from which they then make a subsequent return to their spawning area during the maturation phase (Frank 1992, Brickman 2003) Similar events appear to be a characteristic of many Caribbean fishes (Cowen et al 2006) Discovery of this phenomenon has resulted in the need for significant upward revisions in the estimates of year-class strength at the natal site and a corresponding bonus to the local fishery Its implication for the assessment and management of adjacent stocks is also profound

originat-Paradigm 3: In marine temperate systems, fish spawn in springtime so that peak larval abundance coincides with maximum prey availability (Cushing’s match/mismatch hypothesis)

The Hjort and Cushing hypotheses, which were so instrumental in directing the early research into the population dynamics of marine fishes, were founded on the dynamic interaction between the temporal and spatial distributions of larval fishes and their prey Leggett & DeBlois (1994) system-atically reviewed the scientific evidence for and against these hypotheses and found Hjort’s hypoth-esis to be wanting Support for Cushing’s hypothesis was judged equivocal, mainly because of the difficulty of adequately operationalising the hypothesis and the technical challenges of assembling data on appropriate time and space scales, realities acknowledged by Cushing himself in his update

of the hypothesis (Cushing 1990)

An important feature of the Cushing hypothesis was its removal of the restriction inherent

in Hjort’s thesis that food-mediated mortality would be restricted to a brief ‘critical’ stage in val development (the transition from endogenous to exogenous feeding) Under the Cushing model

lar-recruitment success became associated (at least in theory) with the abundance of food during the entire period of larval development one of the explicit corollaries of this thinking was the expecta-tion that spawning in north temperate marine fishes would be adaptively linked, temporally, to the annual spring (and in some cases fall) production cycle(s) in the sea This corollary also determined and directed much of the research into the link between ocean physics, primary and secondary production, and the recruitment dynamics of fish populations.

Solid evidence of the importance of this link to recruitment has been elusive (Cushing 1990, Leggett & DeBlois 1994) Two papers highlight the contribution of advances in the understanding

of physical/biological linkages in the ocean and the availability of more sophisticated sampling technologies, in particular satellite remote sensing, to an understanding of the potential importance

to the timing of spawning as a determinant of larval survival and recruitment in this species While possibly a case of overinterpretation due to the use of indirect approximations of the independent and dependent variables, the findings are nonetheless supportive of the Cushing hypothesis

In a more direct and convincing study, Platt et al (2003) employed satellite remote sensing to determine the timing of the spring phytoplankton peak on the eastern Scotian Shelf This they related

to peaks in the production of larval haddock While their data series was short (8 yr) and was divided into two distinct periods (1979–1981 and 1997–2001) their analyses revealed a strong relationship

(r2 = 0.89) between variation in recruitment and variation in the timing of the spring bloom.Unresolved by both studies is the extent to which other components of the ecosystem, perhaps themselves linked to and affected by physical attributes of the ecosystem that are co-related to the timing of the spring bloom (predator abundance and diversity, temperature effects, etc.), might influence the environment occupied by eggs and larvae (either positively or negatively) and the extent to which this result can be generalized to other areas and other species

In contrast to these findings, Mousseau et al (1998), who examined the annual cycles of dance of fish larvae and their zooplankton prey in relation to the biomass and production of phyto-plankton on the Scotian Shelf, discovered that the production of copepod nauplii and copepodites was sustained throughout the year and that fish larvae specializing on copepod prey also occurred year-round This occurred notwithstanding the fact that the spring bloom of large phytoplankton, normally assumed to form the basis of the food web for larval fishes, was restricted to February–April This year-round food availability was produced by a non-peak food web structured on the large-microphage shunt of the microbial food web (small phytoplankton → appendicularians/ pteropods → fish larvae) Mousseau et al (1998) concluded that the year-round presence of fish lar-vae and the fact that several of their major prey items exploit the microbial food web challenges the long-standing belief that the feeding of marine fish larvae depends primarily on the reproduction of herbivorous calanoid copepods grazing the spring and autumn blooms of large phytoplankton

abun-De Figueiredo et al (2005) provide further support for this link between larval feeding and the large microphage shunt of the microbial food web They found that Protozoa and appropriately sized metazoan prey, previously largely ignored as potential food items, can contribute significantly to the dietary and energy requirements of larval fish They conclude that the inclusion of these dietary

components into measures of the food resource available to larval fishes diminishes the validity

of the key assumption underlying the Cushing hypotheses — that is, when, at times other than the spring bloom, food levels in the sea are limiting to larval growth and survival

There is mounting evidence, as well, that at larger scales, water temperature may play a greater role in mediating larval growth and survival than food availability (see Meekan et al 2003, Takasuka

& Aoki 2006) This, too, raises the questions, what other components of the ecosystem (predator ties, types, sizes, feeding rates, etc.) are influenced by physical factors linked to the timing of the spring bloom, and how do these ecosystem level interactions influence larval survival and recruitment?

densi-A second corollary of the Cushing hypothesis is that larvae that experience superior feeding conditions, by virtue of their association with areas/times of enhanced food abundance, should exhibit faster growth and higher survival, leading ultimately to improved recruitment This corol-lary, the so-called ‘growth-mortality’ hypothesis (Anderson 1988, Hare & Cowen 1997) is, in fact, comprised of three non-exclusive functional hypotheses: the ‘bigger is better’ hypothesis, the ‘stage duration’ hypothesis and the ‘growth-selective’ hypothesis (Figure 4)

Underlying all three hypotheses is the general assumption that following the transition to enous feeding the risk of death to an individual will be inversely related to the quantity and quality

exog-of food available and, by extension, to rates exog-of growth and development The positive effect exog-of high growth and development rates on survival is generally believed to result from the rapid increase in length and/or changes in behaviour related to development during this period that, in turn, differen-tially influences the susceptibility of individual larvae to predation (reviewed in Litvak & Leggett

1992, Leggett & DeBlois 1994)

‘Bigger is better’ hypothesis

The ‘bigger is better’ hypothesis evolved, in large part, from the results of general models that aggregate data at the species level (an approach criticized by Pepin & Miller (1993) for its distorting effects when generalized to intraspecific studies) and, in part, from the results of laboratory studies that showed larger larvae were less vulnerable to predation However, as demonstrated by Litvak & Leggett (1992) the experimental designs typically used in these studies to assess the relationship between size and vulnerability commonly compounded the effects of age and size by using larvae

of different ages to obtain the desired range of sizes investigated Furthermore, most studies of the effects of size and/or age failed to provide the predator with a choice of prey sizes/ages and there-fore examined only the capture component of the predation act, whereas predation involves three multiplicative probabilities: encounter, attack and capture (reviewed in Litvak & Leggett 1992)

Laboratory and in situ mesocosm experiments in which larvae of identical ages but different sizes

and of identical sizes but different ages were subjected to predation by both visual and non-visual predators clearly illustrated this bias (Litvak & Leggett 1992) Contrary to the predictions of the big-ger is better model, in predation trials involving experimental cohorts of larvae of identical age, but

possessed of a distribution of sizes, both non-visual (jellyfish) and visual (fish) predators consumed disproportionately more larger larvae And, when these predators were offered cohorts of larvae of identical sizes but possessing a distribution of ages, older larvae were selectively consumed This also is contrary to expectation Pepin et al (1992) report similar findings Bertram & Leggett (1994)

found no difference in predation risk when metamorphosing winter flounder (Pseudopleuronectes americanus) of like age but different sizes and of like size but different ages were subjected to

predation by Crangon sp The principal difference in all these experiments, relative to those from

which the bigger is better model was developed, is that the predators were offered a choice of larvae

of different sizes or ages upon which to prey, as would occur in nature

As noted by Leggett & DeBlois (1994), the findings reported by Litvak & Leggett (1992) and

supported by Pepin et al (1992) and Bertram & Leggett (1994) suggest that, contrary to predictions

of the bigger is better model, for individual members of a larval cohort being smaller at a given age could, under some conditions, and at certain stages of development, confer a survival advantage This difference in perspective derives from an explicit recognition of the reality that, in nature, individual larvae exist, and are preyed upon, as members of a population characterized by evolving

distributions of ages, sizes and sizes at age Supporting field evidence is provided by the work of

Fortier & quinonez-Velazquez (1998), who studied the hatch date frequency distributions and daily growth rates of pollock and haddock larvae on the Scotian Shelf in 1992 and 1993 They found that, in haddock, growth varied little over the hatching season and that there was no significant relationship between growth and survival In pollock, slow growth invariably resulted in low sur-vival, but fast growth resulted in both low and high survival of individual daily cohorts Fortier & quinonez-Velazquez (1998) concluded that fast growth was a contributing but insufficient condition for enhanced survival and that for both species investigated, increased predation late in the hatch-ing season could decouple growth and survival It appears, therefore, that differences in individual vulnerability to predation, and the population effects of this variability on recruitment, will be a function of the evolving structure of the distributions of larval size, age, and size at age throughout the larval period and of the predators’ response to them Clearly, many of the inferences initially drawn from the bigger is better model require rethinking

‘Stage duration’ hypothesis

The ‘stage duration’ hypothesis was advanced simultaneously by Houde (1987) and Chambers & Leggett (1987) This model is founded on the observations that (1) variance in sizes at metamorpho-sis in larval fishes is lower than variance in sizes at intermediate ages between hatching and meta-morphosis, thereby implying a ‘target size’ for metamorphosis and (2) the timing of metamorphosis

is more strongly related to the achievement of a target size rather than a target age (Chambers et al 1988) Therefore, larvae that experience more favourable feeding conditions and grow more quickly should develop more rapidly and achieve metamorphosis at earlier ages The stage duration hypoth-esis predicts that, as a consequence, faster-growing larvae should experience lower cumulative mor-tality during the larval stage, during which mortality rates are known to be very high Simulation studies of the potential impact of the reduced time to metamorphosis resulting from accelerated growth indicate that resulting survival could be increased up to 100-fold by rapid growth (Chambers

& Leggett 1987, Houde 1989)

However, when Bertram (1993) incorporated the size-/age-specific mortality relationships tified by Litvak & Leggett (1992) and Pepin et al (1992) into the stage-specific survival model in a manner that recognized the combined probabilities of detection, attack and capture that character-ize vulnerability to predation, he observed that the hypothesized positive effect of fast growth on cumulative survival inherent in the stage-specific model was negated Indeed the revised model demonstrated that, in some cases, survival advantage went to slower-developing individuals This

iden-outcome is consistent with the experimental results reported by Litvak & Leggett (1992) and Pepin

et al (1992), with the finding by Pepin et al (2003) that wild, fast-growing larvae studied via

high-resolution in situ sampling experienced higher rates of mortality than did slower-growing larvae and

with the conclusion by Fortier & quinonez-Velazquez (1998) that rapid growth and development is

a necessary, but not sufficient, condition for improved survival

‘Growth-selective’ hypothesis

The ‘growth-selective’ hypothesis derives principally from the work of Takasuka et al (2003), who

found that slower-growing larvae experienced higher rates of predatory losses independent of their individual size at capture (hence the name ‘growth-selective’ hypothesis) Inherent in this hypoth-esis is the assumption that the nutritional/physiological state of individual larvae may influence their probability of capture (perhaps because of a reduced capacity to detect and/or avoid preda-tory attacks) independent of size (‘bigger is better’) or developmental state (‘stage duration’) Their

in situ analysis was conducted by comparing the size and growth rates of individual larvae ered from the guts of a suite of predators (as assessed through otolith analysis) with the distribution

recov-of sizes in the larval prey field exploited by these predators Significantly, they found evidence

of both negative and positive size-selective predation depending on the particular predator cies investigated However, the combined effect of all predators was neutral This finding helps to explain the conflicting results of individual studies in which one or a small number of predators was employed or studied

spe-A follow-up study (Takasuka et al 2004) provided further evidence of predator-specific

growth-selective mortality Larval anchovy (Engraulis japonicus) having slower recent growth rates than

those in the population from which they were consumed (independent of size at capture), exhibited higher predation losses when preyed on by juvenile conspecifics However, predation by skipjack

tuna (Katsuwonus pelamis) on the same population of anchovy larvae produced no such

growth-selective mortality

Fuiman et al (2006) examined the presumed basis for growth-selective predation in a suite of

experiments involving larval red drum (Sciaenops ocellatus) These experiments were conducted both in the laboratory and in in situ enclosures They found that populations of larval red drum

consisting of 15 fast- and 15 slow-growing larvae of comparable size, when exposed to predation

and maintained in in situ enclosures, showed no evidence of growth-selective predation This result

could, however, be a product of the particular predator used and is consistent with the findings of Takasuka et al (2004) Laboratory experiments involving larval red drum designed specifically to evaluate the behavioural dynamics presumed to underlie growth-selective predation revealed no evidence of a growth rate effect on performance in 11 survival skills This outcome raises important questions about the fundamental assumption of the growth-selective hypothesis

The strong focus of research on the importance of size, growth rate and development rate as determinants of survival reflected in these three hypotheses, and in the creative field and laboratory experiments designed to assess them, has meaningfully advanced knowledge of the behaviour, ecol-ogy and dynamics of the larval stage of fishes To date, however, it has contributed only modestly

to the ultimate objective of a fuller understanding of recruitment processes in fishes Like the Hjort and Cushing hypotheses that preceded them, there may be a danger that this focus, if sustained at the expense of other research directions could stall advances in the field by precluding a more holis-tic view of the factors regulating larval survival and recruitment The hypothesis that larger larvae should experience a survival advantage when subjected to predation is, for example, counter to the predator-based hypothesis of optimal foraging that predicts predators will actively select larger prey items from a prey field consisting of individuals of different sizes (Werner & Hall 1974, Charnov

1976, Litvak & Leggett 1992) Moreover, the underlying premise that the processes inherent in

these growth-related hypotheses are vital to recruitment continues to be built, in the main, on a shaky foundation of correlation studies in which cause and effect are often conflated and/or con-fused For example Jenkins & King (2006) conclude, on the basis of correlation studies, that larval growth in seagrass-associated fishes was important to post-larval abundance and interannual varia-tion in recruitment However, larval growth was also strongly correlated to water temperature, and water temperature was, in turn, strongly correlated with recruitment An equally credible conclu-sion might be that temperature had an influence, either directly or indirectly (see Frank & Leggett

1982, 1983), on survival independent of growth rates In fairness to Jenkins & King (2006) some consideration was given to other processes, including variable physical transport processes that affected larvae differentially from year to year (of which variable sea-surface temperature could be

an indicator) However in this example, as in others (see, for example, Meekan et al 2003, Takasuka

& Aoki 2006), the link between increased growth rates and enhanced survival and recruitment has become almost axiomatic

Paradigm 4: Environmentally based recruitment models, when updated with new data, invariably fail

There is a long history of attempts to explain variation in recruitment based on the relationship between some direct or indirect measure of year-class strength and environmental variables (Cushing 1982) The most commonly used environmental variables have been temperature, salin-ity and wind These have been incorporated into exploratory correlation analyses that typically use indices of year-class strength based on landings, catch per unit of effort, fishery-independent scientific surveys, or model outputs based on reconstructed population sizes such as virtual popula-tion analysis (VPA), as measures of recruitment Unfortunately, estimation of year-class strengths using these approaches is not a trivial problem and the resulting estimates can be misleading often researchers have been forced to use commercial landings data or landings per unit of effort (e.g., Sutcliffe et al 1977), data sources that typically fail to conform to the key assumption of time-invariant fishing effort Such data series have been distorted by technological developments that change the relationship between a unit of effort and the resultant catch More reliable data, derived from fishery-independent surveys, are, unfortunately, scarce

Temperature, because it regulates many physiological processes, has long been considered an important explanatory variable of recruitment and has recently received growing attention in the context of global warming (Cardinale & Hjelm 2006) one general pattern that has emerged is that above-average temperatures are associated with better survival among stocks occupying the northern limit of their distribution (ottersen & Stenseth 2001) Salinity has frequently been used

as an indirect measure of nutrient flux, with higher-than-average salinity corresponding to elevated nutrients Sutcliffe et al (1983) found a positive correlation between cod recruitment and salinity

on the Newfoundland/Labrador shelves, which they assumed represented increased vertical mixing that supported increased plankton production The physical process by which wind may influence recruitment is thought to be primarily through distributional effects on the egg and larval stages (but wind-induced upwelling could also increase vertical mixing and plankton production) For

example, the larval stages of Pacific hake (Merluccius productus) appear to be negatively affected

by along-shelf winds that induce offshore Ekman transport to areas where survival is poor (Bailey

1981) Conversely, Atlantic menhaden (Brevoortia tyrannus) appear to benefit from both alongshore

and longshore wind-induced transport to coastal nursery grounds (Stegmann et al 1999) While there is sufficient biological justification to consider these physical variables as important contribu-tors to recruitment variability it should be noted that they are often selected due to their ease of measurement and the long record of measurement — decades to centuries in some cases

Unfortunately, this rationale makes it unlikely that such relationships will withstand the test

of time Indeed, when updated with new data such relationships are prone to failure In a hensive reexamination of over 60 published environment-recruitment correlation studies of marine fish and invertebrates, Myers (1998) tested the performance of each using data accumulated sub-sequent to the publication of the original relationship In many cases the analysis applied to these new data duplicated exactly the original The overall result was disappointing Most previously published relationships failed when updated with new data Interestingly, those correlations that remained significant when retested generally occurred in populations close to the limits of the spe-cies range In northern waters this included Barents Sea cod, Baltic Sea cod, and Pacific herring

compre-(Clupea pallasii) Recruitment in these species exhibited a consistently positive relationship with

temperature At the southern limit of the range limits recruitment was typically negatively lated with temperature

corre-Insights into the possible causes of such latitudinally linked effects are provided by the work

of Fogarty et al (2001), who examined the relative variability in recruitment of cod and haddock stocks distributed throughout the north Atlantic Recruitment variability was measured as the stan-dard deviation of the residuals from a Ricker stock and recruitment relationship Despite their close taxonomic relationships, overlapping spawning times, and similar egg sizes, haddock stocks gen-erally exhibited higher recruitment variability than cod in each of nine geographic areas where they co-occurred Both stocks also exhibited higher variability at the extremes of their geographic ranges (Barents Sea in the north and Georges Bank to the south; Figure 5)

The authors explain these species-specific differences on the basis of contrasting life-history traits, physiological tolerances and population dynamic processes For example, cod, which spawn over a more extended time period than haddock, may thereby dampen the risk of extreme survival outcomes (boom-bust) due to environmental variability Laboratory studies (Laurence 1977) show that cod larvae can withstand a much broader range of temperature and salinities than haddock, implying that equivalent environmental forcing can be expected to have a lesser impact on growth and survival of larvae of cod relative to haddock The reduced temporal variability in recruitment

of cod relative to haddock also led Fogarty et al (2001) to speculate that density-dependent nisms, possibly cannibalism, may be a more important process in cod These species differences

ea

Iceland

W S

cotian E.

cotian

Georg

Figure 5 Variability in the long time series of recruitment data for cod and haddock showing that in every

region where the two species co-occur, haddock exhibit higher interannual variability in recruitment, tive of stronger environmental effects on haddock than on cod.

sugges-have important implications for the management and potential recovery timescales of the two cies For example, the formation of large year-classes at low levels of spawning stock may be more common in haddock, thus serving to initiate the process of stock recovery Evidence in support of this possibility may be drawn from the study by Platt et al (2003) of haddock larval survival in rela-tion to spring bloom timing and the apparent recovery of the eastern Scotian Shelf haddock stock following its collapse in the early 1990s (Figure 6).

spe-Myers (1998) offers several statistical and analytical suggestions for improving the success of future attempts to link environmental factors to recruitment variability These include testing of general hypotheses through the examination of several populations at once to detect general pat-terns; employing data-splitting procedures — one portion for the exploratory analysis and the other for testing; correcting for variation in spawner abundance (trends in recruitment may result from changes in SSB)

Notwithstanding these inherent difficulties, such correlative approaches remain important tools for identifying relationships worthy of more in-depth investigation through a combination

of laboratory and in situ studies and simulation modelling Ultimately, however, the development

of recruitment models with ‘staying power’ will likely depend upon the acquisition of a greater understanding of the biology of the species and of the behaviour of key elements of the physical

and biological environment occupied in support of the development and rigorous testing of a priori

hypotheses This hypothesis-testing approach has been more common in freshwater than marine systems (Magnuson 1988, Frank & Leggett 1994), in part because of the many logistic advantages

of working at smaller scales and in what are often less complex systems

one of the few examples from the marine environment of a long-lasting ment model involves the demonstrated link between the frequency of onshore winds during the period of residence of the larvae of capelin in the beach gravel where they are spawned and their survival and recruitment (Leggett et al 1984) This model rests on the results of an extensive series

environment/recruit-of field and laboratory investigations focused on the early life history environment/recruit-of the species in eastern Newfoundland waters In the case of capelin, onshore winds produce a shoreward movement of warmer, prey-rich waters that displace the cold, upwelled, predator-laden waters that typically dom-inate the near shore during the capelin spawning season (Frank & Leggett 1982) A combination of passive (wave-induced) and active (apparently temperature-induced) emergence from the spawning gravel follows hatching These cause the larvae to become entrained in a water mass rich in prey and relatively poor in predators (Frank & Leggett 1982, 1985, Figure 7), thereby increasing the survival probabilities of individual larvae (Frank & Leggett 1983) The frequency of these onshore wind events and of the resulting emergence events is also key because survival time of larvae in the gravel

is limited to 3–4 days post-hatch — a time limit set by the rate of consumption of yolk reserves

0

1970 1975 1980 1985 1990 1995 2000 2005 20

Figure 6 Spawning stock biomass (dark line), fishable biomass (dashed line) and exploitation rates (grey line)

for eastern Scotian Shelf haddock.

following hatching In years when the frequency of onshore winds is high, survival is favoured When onshore winds are infrequent, survival is depressed.

During the period of larval hatching, the combined effects of water mass replacement ing favourable conditions for feeding, growth and survival coupled with short post-hatching beach resident times result in increased larval survival and recruitment The original model documenting this relationship was based on capelin year-class data from 1966 to 1987 and from wind field data that reflect variability in the atmospheric pressure systems that generate onshore winds over large areas of Newfoundland’s east coast It explained nearly 60% of the interannual variation in year-class strength in the eastern Newfoundland capelin stock (Leggett et al 1984) Higher-than-average water temperatures during the 6-month period of post-emergent pelagic drift also had a positive effect on year-class strength There is no documented biological basis for this added contribution

creat-to explained variance When retested with an additional 16 yr of data the model was found creat-to be robust (Carscadden et al 2000) However, the added contribution to explained variance contrib-uted by the incorporation of temperature during the post-emergent drift interval into the model no longer prevailed — another example of a failed correlation based on purely exploratory correlative assessments This body of research provides an important counter-example to the paradigm that

0

1970 Quiescent Capelin Emerging Capelin

Figure 7 Dynamics of the nearshore environment influencing the larval stages of capelin and the resulting

impact on recruitment The histogram shows onshore wind interval.