The future of fisheries oceanography lies in the pursuit of multiple hypotheses

Abstract

Fisheries oceanography is largely an applied discipline with a major goal of improving fisheries management and marine conservation. Johan Hjort's critical period hypothesis, and its decedents, remain a dominant theme and focuses on year-class success as mediated by prey availability and feeding. Bottom-up forcing, a related hypothesis, focuses on the sequential transfer of energy through the pelagic foodweb from primary productivity to fishery productivity. Another approach assumes that trophic interactions of adults determine abundance. Fisheries assessment and management, however, is based on the hypothesis that fishery abundance is determined by time-varying fishing and year-class success related to spawning-stock biomass. These approaches, their basic hypotheses, and underlying processes and mechanisms suggest very different dynamics for fishery populations. Other hypotheses challenge these traditional views: predation of early life stages, parental condition, shifting migration pathways, and physiological limits. Support for these other hypotheses is reviewed and the research needs are described to apply these hypotheses to fisheries assessment and management. Some of these hypotheses were identified by Hjort (e.g. parental condition hypothesis) and others are relative new (e.g. early life stage predation hypothesis). Moving into the future, we should focus on Hjort's approach: multi-hypothesis, integrative, and interdisciplinary. A range of hypotheses should be pursued with an emphasis on comparing and linking multiple hypotheses. The results then must be incorporated into fishery assessments and management decisions to support the long-term sustainability of exploited species and the conservation of threatened and endangered species.

Introduction

Fisheries oceanography defined

Fisheries oceanography is defined as the study of oceanic processes that affect marine ecosystems and the relationship of these ecosystems to the abundance, distribution, and availability of fishery species (Harrison and Parsons, 2000; other definitions are summarized in Kendall and Duker, 1998). The emphasis is on exploited species, but fisheries oceanography includes species and processes that affect and are affected by exploitation. Species of interest include phytoplankton, zooplankton, finfish, shellfish, marine mammals, and sea turtles; processes of interest encompass the fields of oceanography, fisheries, and ecology. This essay treats fishery abundance and distribution. Availability to a fishery remains a relatively understudied facet of fisheries oceanography and involves an understanding of the interaction among the marine ecosystem, fishery populations, and fishing activities (e.g. gear-selectivity, socio-economics, fishing technology, regulation, etc.).

Basically, fisheries oceanography is an applied science with the potential to inform scientists, managers, and fishers. Many early researchers explicitly identified the applied aspects of their work as a main motivating factor (Hjort, 1914, 1926; Walford, 1938; Sette, 1943). Although “improving fisheries management” is stated as a goal in many fisheries oceanography studies, the direct connection to management is not often made. However, recent advances have brought us to the brink of widespread application of fisheries oceanography to fishery assessment and management (Köster et al., 2003a, b). Also, the recognition of the need to manage fisheries in the context of the broader ecosystem (e.g. Ecosystem-Based Fisheries Management) brings fisheries oceanography to the forefront as a requirement for assessment and management (Rothschild, 1986; EPAP, 1999; Pikitch et al., 2004; Cury et al., 2008). In this essay, I track the development of fisheries oceanography, discuss the link between hypotheses and fisheries assessment, and suggest future directions for the field. In honour of Hjort's centennial, I provide an overview of the critical role of Hjort's hypotheses in the development of fisheries oceanography and advocate that we must embrace his approach, in addition to his hypotheses, to achieve the basic goal of improving fisheries assessment and management.

Johan Hjort sets the stage

Arguably, Johan Hjort is the father of fisheries oceanography. His 1914 and 1926 publications set the stage for a century of work aimed at understanding fluctuations in abundance of fishery species. Hjort knew that fishery yields varied through time, and his purpose was to understand the basis for this variability. He saw the potential to help fishers understand and respond to the natural dynamics of the resource and he even wrote about developing a predictive capability. His approach was hypothesis-driven (Table 1), and he brought a number of lines of evidence and tools to bear in his investigations. He also synthesized information from other researchers and from several stocks of both Atlantic cod and Atlantic herring. Hjort did not work in a vacuum and an excellent summary of the historical and scientific context of his work is provided by Sinclair (1997).

Table 1.

Definitions of hypothesis, process, and mechanism as used in this essay.

Hypothesis: A proposed explanation for a phenomenon that can be tested. In this essay, the phenomenon is variability in the abundance and distribution of fishery species

Process: An ecological action that reaches a particular effect. Five processes are identified here as causing variability in the abundance and distribution of fishery species: (i) prey/feeding, (ii) predation, (iii) movement/dispersal, (iv) ecophysiology, and (v) reproduction/spawning

Mechanism: A system of causally interacting parts that constitute a process in a specific case. For example, the mechanism of the parental condition hypothesis proposed by Friedland et al. (2008) consists of seven parts (Figure 3)

In total, Hjort considered two general hypotheses and three subhypotheses (Table 2). When Hjort began his work, the standing hypothesis was that variable fishery yields were caused by variable adult migration (H₁), which led to variable overlap with fishing effort; abundance varied relatively little (see Sinclair, 1997). Migrations were caused by seasonal changes in the environment (abiotic and biotic) and the sensitivities of adult fish to these changes. Through the investigations of Hjort and his contemporaries, the migration hypothesis was discounted. In its place, Hjort accumulated a large amount of evidence to support the hypothesis that fluctuations in fishery yields resulted from fluctuations in population abundance caused by variability in year-class success (H₂), defined as the number of individuals that survive early life stages (eggs, larvae, and juveniles) to join the adult population. Hjort acknowledged that the renewal of a fish stock was dependent on many factors and that poor conditions in any one factor could determine the ultimate result. However, he identified and discussed three subhypotheses regarding the determination of year-class success: (a) quality and quantity of spawning (H_2a), (b) nourishment of young larvae (H_2b), and (c) passive movement of larvae in currents (H_2c) (often termed Hjort's second hypotheses, Sinclair et al., 1985). Based on examination of cod liver and roe, Hjort found the years with highest year-class success were correlated with the lowest roe weights and thus, he discounted the quality and quantity of spawning hypothesis (currently termed the parental condition hypothesis). With regard to the nourishment of larvae, he discussed the potential importance of starvation after hatch (the critical period hypothesis) and the timing of spawning relative to the timing of primary production (currently termed the match–mismatch hypothesis). He did not discuss the dispersal of larvae in detail but recognized the potential importance of passive and active movements to and from favourable conditions.

Table 2.

List of hypotheses considered here.

Hypothesis	Source	Description
1. Migration/movement	Hjort (1914)	Variability in migration pathways result in variability in availability to fisheries while adult abundance remains relatively constant; distribution varies as a result of variable movement (the stipulation that abundance is relatively constant is no longer part of the migration/movement hypothesis)
2. Year-class success	Hjort (1914)	Variability in survival during prerecruit stages result in variability in adult abundance
2a. Parental condition	Hjort (1914)	Variability in parental investment in offspring results in variability in prerecruit survival
2b. Prey environment	Hjort (1914)	Variability in prey abundance and prey distribution results in variability in prerecruit survival
2c. Dispersal	Hjort (1914)	Variability in dispersal results in variability in prerecruit survival
2d. Predation	Bailey and Houde (1989)	Variability in predation results in variability in prerecruit survival
2e. Ecophysiology	Fry (1971), Neill et al. (1994)	Variability in the physiological environment results in variability in prerecruit survival
2f. Growth-mortality	Anderson (1988)	Variability in the interaction between growth and mortality results in variability in prerecruit survival
3. Fishing and recruitment	Beverton and Holt (1957)	Variability in survival during prerecruit stages and fishing mortality result in variability in adult abundance
4. Adult trophic interactions	Odum and Odum (1955), Polovina (1984)	Variability in adult prey and predation results in variability in adult abundance

The four general hypotheses are denoted by Arabic numerals and the six subhypotheses are denoted by letters. This notation is used throughout the text. This list is not comprehensive in terms of hypotheses that explain variability in the abundance and distribution of fishery species (see Figure 2 and Houde, 2008).

Since Hjort's publications, his first-feeding hypothesis (H_2b) received the most attention. From this research, many of the classic tenets of fisheries oceanography originated. The critical period hypothesis proposes that year-class success is determined by prey availability at the transition from yolk-sac to exogenous feeding (Hjort, 1914). The match–mismatch hypothesis proposes that year-class strength is determined by the match between prey production and spawning cycles of the focal species (Cushing, 1990). The stable ocean hypothesis proposes that ocean stability allows aggregation of larval prey, thereby enhancing feeding conditions and by extension year-class success (Lasker, 1978). Similarly, the optimal environmental window hypothesis states that year-class success is related to local optimal environmental conditions that maximize prey production (Cury and Roy, 1989). In general, these hypotheses are refinements of Hjort's original hypothesis and propose that the feeding environment of planktonic larvae is the dominant factor shaping year-class success and ultimately the abundance of fishery populations. There are examples that provide strong support for these prey-related hypotheses (e.g. Ringuette et al., 2002; Castonguay et al., 2008), but the support is not universal (May, 1974; Houde, 2008).

Over the past century, efforts to understand fluctuations in fishery abundance have not only focused on the larval feeding environment. Walford (1938) examined the retention of haddock larvae on George's Bank. This work followed Hjort's dispersal hypothesis (H_2c) and many others have examined this hypothesis (e.g. following Walford: Colton and Temple, 1961; Myers and Drinkwater, 1989; Boucher et al., 2013). The role of larval dispersal in broader population dynamics was further developed in the member-vagrant hypothesis (Sinclair, 1988) and through studies of population connectivity (Pineda et al., 2007; Cowen and Sponaugle, 2009). Like the larval feeding hypotheses, the larval dispersal hypothesis proposes that year-class success is a dominant component of fishery abundance but is determined by dispersal to favourable habitats for continuation of the life cycle rather than the larval feeding environment.

Towards forecasts of fishery abundance and distribution

Taking a step back from Hjort, four general hypotheses exist that explain fluctuations in fishery abundance (Figure 1 and Table 2). The first two hypotheses were dealt with by Hjort: movement/migration (H₁) and year-class success (H₂). The third is the standard fisheries hypothesis (H₃): recruitment is related to spawning-stock biomass and interannual changes in fishing mortality are responsible for changes in total mortality (Hilborn and Walters, 1992; Quinn and Deriso, 1999). The role of overfishing in declining fishery abundance was investigated during Hjort's time, but by a different group of scientists (see Sinclair, 1997). The fishing hypothesis assumes that year-class success is an important driver of fishery abundance (sensu Hjort) but does not try to understand recruitment variability beyond that explained by spawning-stock biomass; this variability is modelled simply as randomly distributed error in the stock–recruitment relationship. A population can be managed at a certain level, if a certain fishing mortality can be maintained. Recruitment can be measured by surveys, thus removing the need to understand the processes that cause variability in recruitment. The fourth hypothesis (H₄) is based on the transfer of energy through foodwebs (Odum and Odum, 1955; Polovina, 1984). Biomass of a given component of an ecosystem depends on food resources and predation pressures. This approach originated from ecological theory and has expanded greatly in the study of marine systems, which have been represented as food chains (e.g. Ryther, 1969) and foodwebs (e.g. Pauly et al., 2000). This hypothesis focuses on trophic interactions of adult stages and ignores the complex life history of many marine species.

Figure 1.

A schematic illustrating the four main hypotheses for variability in fishery abundance: (1) migration, (2) year-class success, (3) fishing and stock–recruitment relationship, and (4) adult trophic interactions. See Table 2 for a brief description of these hypotheses. The colours correspond to the main operating process of the hypothesis (Figure 2).

To further develop the applied aspects of fisheries oceanography, the differences between these four general hypotheses need to be recognized and links need to be made among their implied processes and mechanisms (Figure 1, Table 1). For example, over the past 25 years, a large amount of effort has gone into modelling the physical environment, primary productivity, and zooplankton productivity. Many of these models have been linked to fishery populations through larval feeding and larval dispersal (Werner et al., 2001, 2007; North et al., 2009; Hinrichsen et al., 2011). At the heart of these models, often implicitly, is the hypothesis that variability in year-class success drives variability in population abundance (H₂). Further, variability in year-class success is determined by the larval feeding environment (H_2b) or larval dispersal (H_2c). Thus, these efforts are based on the classic hypotheses of recruitment fisheries oceanography (sensuKendall and Duker, 1998). An important limitation to applying these oceanographic models is that they consider only a subset of the hypotheses that explain fishery abundance and distribution.

The issue of developing forecast and predictions based on narrowly focused models is not restricted to the investigation of year-class success. Similarly, limited examples could be provided for the remaining three general hypotheses identified as controlling fishery abundance (H₁, H₃, H₄). Fisheries oceanography, as defined above, is compartmentalized in terms of the hypotheses that are used to explain variability in fishery abundance and distribution. Oceanographers and ecologists focus on variability in year-class success (H₂), fishery assessment scientists emphasize the stock–recruitment relationship and the effect of fishing (H₃), and ecosystem scientists concentrate on trophic interactions of adults (H₄). One purpose of this essay is to articulate this compartmentalization and to stimulate thought and effort to work across these scientific communities and to integrate across the four general hypotheses explaining variability in fisheries.

Pursuing hypotheses

Through examination of the work of Hjort, the subsequent 100 years of fisheries oceanography research, and the four general hypotheses presented above, five regulating processes can be identified: (i) prey, (ii) predation, (iii) movement, (iv) ecophysiology, and (v) reproduction. Reproduction is the start of the life cycle and involves a great deal of complexity (see Cole, 2010). Most marine organisms have complex life histories, with different ontogenetic stages occupying different habitats from spawning to maturity (Figure 2), and the remaining four regulating processes act during all life stages. There are interactions among all five processes resulting in the complex dynamics of fishery abundance and distribution. The ideas of prey, predation, movement, and reproduction are introduced above. Ecophysiology is the interrelationship between the abiotic and biotic environment and an organism's physiology (Neill et al., 1994).

Figure 2.

A schematic showing the different life stages of the typical marine fishery species and the five general processes that affect abundance and distribution: (i) prey/feeding, (ii) predation, (iii) movement/dispersal, (iv) ecophysiology, and (v) reproduction/spawning. Potentially, any process during any life stage could contribute to variability in the abundance and distribution of fishery populations. The hypotheses discussed here are indicated by solid arrows at relevant life stage and process: (H₁) migration, (H₂) year-class success, (H_2a) parental condition, (H_2b) prey environment, (H_2c) dispersal, (H_2d) predation, (H_2e) ecophysiology, (H_2f) growth-mortality, (H₃) fishing and stock–recruitment, and (H₄) adult trophic interactions. The numeric notation used throughout the text and the hypothesis are briefly described in Table 2. Unfilled arrows are life stage and process combinations that are not included here. All of these life stages and processes likely act to affect abundance and distribution of fishery species; however, the rank importance of each varies over species, populations, space, and time. There is also a large degree of interaction among processes requiring a collaborative, interdisciplinary, synthetic, multi-model approach.

Viewing specific hypotheses and processes within the set of possible hypotheses and processes highlights the limited nature of much of our work and illuminates new areas for research (Figure 2). To achieve the goal of improving fisheries assessment and management, multiple hypotheses must be examined in more detail and incorporated into assessment models. In addition, the hypotheses not considered in an application of fisheries oceanography must at least be acknowledged. Here I discuss four hypotheses with examples from the Northeast US Shelf Large Marine Ecosystem. Two of these hypotheses were dealt with by Hjort [parental condition hypothesis (H_2a) and migration hypothesis (H₁)], but for whatever reason, are relatively understudied compared with the critical period hypothesis (H_2b) and the larval dispersal hypothesis (H_2c). The other two hypotheses were developed in the latter half of the 20th century: predation on early life stages (H_2d) and the ecophysiology of early life stages (H_2e).

The four hypotheses discussed here (H_2a, H_2d, H₁, H_2e) imply different dynamics than the traditional hypotheses (H_2b, H_2c, H₃, H₄) and their pursuit is important to understand the consequences of different regulating processes on fishery abundance and distributions. I choose to highlight these hypotheses because I think there is value in their examination: value in developing a scientific understanding of the hypotheses' regulating processes, as well as value in including these hypotheses in fishery assessment and management. Other researchers in other regions would likely choose different hypotheses. My goal is not to produce a review of fisheries oceanography nor to address all extant hypotheses related to variability in fishery populations (see Houde, 2008, for a more complete treatment). Rather I hope to stimulate thought, discussion, research, and application based on hypotheses that I feel are most deserving of pursuit. I use the term pursuit purposely. We need to actively investigate a range of hypotheses related to the dynamics of fishery populations and integrate and apply these investigations to fisheries assessment and management, no matter where the science leads us.

Parental condition hypothesis (H_2a)

The parental condition hypothesis (H_2a) was considered by Hjort (1914). He examined liver weight and year-class success in Atlantic cod and found no relationship. The hypothesis was also acknowledged by Beverton and Holt (1957) in their seminal work on the stock–recruitment relationship and the importance of fishing mortality in fishery dynamics (H₃). Despite these early treatments, the parental condition hypothesis did not become a focus of researchers until the late 1980s–early 1990s (see review by Green, 2008). Both maternal and paternal effects are documented (Green and McCormick, 2005; Trippel et al., 2005; Probst et al., 2006). Increased parental size and weight result in larger and higher condition offspring (Hislop, 1988; Berkeley et al., 2004). Increased parental condition also leads to larger and higher condition offspring (Blanchard et al., 2003; Donelson et al., 2008). Social conditions, such as crowding, influence offspring size and condition (McCormick, 2006). Additionally, evidence for skip spawning (failure to start egg production or resorption of eggs before spawning) has been found in many stocks and linked to poor maternal condition (Rideout et al., 2000, 2005; Skjæraasen et al., 2012). Finally, the environment experienced by parents can modify the phenotype of offspring, thus one life stage can affect another (Salinas and Munch, 2012).

Parental age, size, condition, and environment affect offspring size, growth, condition, and phenotype. Couple these observations with the growth-mortality hypothesis (H_2f, Anderson, 1988) and the hypothesis that parental condition influences year-class success is complete. The growth-mortality hypothesis (H_2f) holds that larger and faster growing larvae have a higher probability of survival and implicitly combines the prey environment hypothesis (H_2b) and the predation hypothesis (H_2d) (Meekan and Fortier, 1996; Hare and Cowen, 1997; Takasuka et al., 2003). By inference, higher parental condition is linked to larger size, higher growth, and increased survival of early life stages, which leads to higher year-class success (e.g. Vigliola and Meekan 2002). Higher larval growth and condition, however, are not a function of the prey environment (H_2b), but due to the parental contribution to offspring (H_2a).

The ideas and concepts of the parental condition hypothesis are central to recent work on the recruitment dynamics of Georges Bank haddock. Friedland et al. (2008) described a significant correlation between the magnitude of autumn bloom and haddock recruitment. Haddock primarily spawn in spring, so autumn bloom is timed with the adults preparing for spawning. A number of other variables were examined (e.g. spring zooplankton, spring advection) and the correlations with recruitment were minimal providing little support for the hypotheses that prey (H_2b) and dispersal (H_2c) influence haddock recruitment. Friedland et al. (2008) proposed that a large autumn bloom results in increased export of material to the bottom, which improves feeding conditions for adult haddock (Figure 3). Increased parental condition results in higher quantity and quality of offspring, which leads to increased year-class success. Previous studies on haddock found parental effects on offspring size and survival (Trippel and Neil, 2004) and significant correlations between recruitment and indices of adult growth and condition (Marshall and Frank, 1999), providing ancillary support for the hypothesis of Friedland et al. (2008).

Figure 3.

The relationship between the magnitude of autumn bloom and the survivor ratio (recruits per spawner) for Georges Bank haddock. The hypothesized mechanism leading to the correlation is a large autumn bloom improves the feeding conditions for adult haddock and results in a greater quality and quantity of offspring and ultimately greater year-class success and higher recruitment (modified from Friedland et al., 2008). The mechanistic steps in the hypothesis from autumn bloom to recruitment are indicated by the boxes on the right side.

The parental condition hypothesis must be further evaluated. In the context of Friedland et al. (2008), there are a number of mechanistic links that must be studied: the conditions that lead to improved parental condition; the physiology of investment into reproduction vs. growth vs. activity; the export of energy from plankton blooms to the benthos; and ultimately, the relationship between parental condition and year-class success. The growth-mortality hypothesis provides a mechanism linking offspring condition to increased larval survival, and this hypothesis must also be tested to evaluate the mechanisms by which larger, higher condition, and faster growing early life stages result in higher recruitment. Many of these scientific questions are outside of traditional recruitment fisheries oceanography as defined over the past 30 years (Kendall and Ducker, 1998), but are directly relevant to understanding the mechanisms that affect fisheries abundance relative to the parental condition hypothesis.

The parental condition hypothesis also needs to be linked with fishery population dynamic models. This hypothesis indicates that the paradigm of a spawning-stock biomass–recruitment relationship is flawed (sensuRothschild and Fogarty, 1989); the emphasis should be on the quantity and quality of offspring produced not on the combined weight of the spawning population. Major steps in this direction have been taken. Marshall et al. (1999) showed that spawning-stock biomass is not necessarily a robust measure of reproductive output (see also Marshall et al., 2003). Further, the hypothesis implies links between reproduction, growth (prey processes), and survival (predation processes).

Parental factors have been included in population models demonstrating that the hypothesis can be incorporated into stock assessments and management advice (Óskarsson and Taggart, 2010; Shelton et al., 2012). Murawski et al. (2001) included parental effects in a stock assessment model for Atlantic cod, specifically linking early life stage survival to the age of spawners. Their results indicate that models not including parental effects may overestimate the resiliency of a population to fishing. Another potential approach is to include the environmental effects influencing parental condition in an environmentally explicit stock–recruitment relationship (see Iles and Beverton, 1998; Köster et al., 2003b). The implications of the two different approaches (modify spawning-stock biomass or model recruitment with additional terms) need to be evaluated, but the parental condition hypothesis can be used in fisheries assessment and management if the necessary information is gathered (specific mechanism, links to models).

Predation on early life stages (H_2d)

Predation has a clear influence on population abundance but was not considered by Hjort (1914, 1926). Fishing, which is human predation, is the dominant source of mortality for many exploited species and directly affects abundance (Myers et al., 1997; Christensen et al., 2003). Many traditional stock assessment models (H₃) assume a constant natural mortality rate, which largely represents non-human predation mortality on post-recruitment stages. The abundance estimates resulting from these models, however, are sensitive to the magnitude and variability of natural mortality (Jiao et al., 2012). Multispecies (Curti et al., 2013) and foodweb models (Overholtz and Link, 2009) also include predation (or consumption) as a dominant regulating process. Traditional stock assessment methods, multispecies models, and foodweb models hypothesize the importance of predation on post-recruitment stages (H₃, H₄), yet Hjort (1914, 1926) hypothesized that fluctuations in fishery abundance were largely determined during prerecruitment stages (H₂). The difference between these hypotheses (adult predation vs. year-class success) is rarely considered.

Interest in predation during prerecruit stages increased in the late 1980s (H_2d, Bailey and Houde, 1989; Verity and Smetacek, 1996; Bailey and Duffy-Anderson, 2010). Studies indicate that predation on early juveniles may be a major driver of abundance in temperate and coral reef ecosystems (Carr and Hixon, 1995; Tupper and Boutilier, 1997). Takasuka et al. (2003), in an elegant study, showed that predation was higher on slower growing larvae, providing direct evidence for an inverse relationship between larval growth and predation (H_2f) and demonstrating an interaction between the processes that affect growth (prey) and those that affect survival (predation).

Two recent lines of investigation further emphasize the role of predation on early life stages in determining year-class success and ultimately adult abundance. First, predation on groundfish larvae by pelagic fish adults has been proposed as a mechanism for low groundfish recruitment and for keeping groundfish populations at low levels even after decreases in fishing effort (e.g. Köster and Möllmann, 2000). Modelling shows that predation on groundfish larvae can be important to population dynamics (Collie et al., 2013) and field studies have found spatial and temporal overlap between groundfish larvae and pelagic fish adults (Garrison et al., 2000, 2002). Second, predation on benthic eggs by groundfish predators has been hypothesized to regulate Atlantic herring abundance on the northeast US continental shelf (Richardson et al., 2011). A model incorporating Atlantic herring spawning-stock biomass and egg predation by haddock predicts the abundance of larvae on Georges Bank over the past four decades (Figure 4). This model is supported by feeding studies that found haddock with large proportions of fish eggs in their stomachs (Langton and Bowman, 1981; Toresen, 1991). In the western Baltic Sea, Polte et al. (2014) found a significant correlation between Atlantic herring yolk-sac larvae and interannual variability in year-class success. They concluded that egg mortalities are a likely cause for recruitment variability. These predation hypotheses remain controversial in part because time-series analyses have not identified a link between groundfish and pelagic fish dynamics (Liu et al., 2012; Swain and Mohn, 2012).

Figure 4.

Time-series of Georges Bank/Gulf of Maine Atlantic herring spawning-stock biomass, larval index, haddock predation intensity, and estimated egg survival. The observed larval index (black line) was compared with a modelled larval index (grey line). The model assumes egg production is proportionate to spawning-stock biomass and then applies haddock predation to the egg biomass where haddock predation is estimated by numbers-at-age and consumption-at-age derived from survey and field studies. Through time, variable haddock predation results in variable Atlantic herring egg survival and the number of larvae estimated by the model closely matches the number of larvae observed in surveys (modified from Richardson et al., 2011). The mechanistic steps in the hypothesis from Atlantic herring spawning to recruitment are indicated by the boxes on the right side.

The predation hypothesis (H_2d) needs to be investigated further. The ability to model larval feeding is quite advanced (e.g. Fiksen and MacKenzie, 2002), yet the ability to model predation on eggs and larvae is relatively rudimentary (but see Fuiman et al., 2006). Most biophysical numerical models do not explicitly include predation. Predation is often implicitly assumed to be constant over time and space (e.g. Cowen et al., 2000) or based on size- or growth-specific mortality functions [e.g. the growth-mortality hypothesis (H_2f), Hermann et al., 2001; Kristiansen et al., 2011]. Recent biophysical modelling studies demonstrate the potential importance of including time and space varying predation (e.g. North et al., 2009; Ji et al., 2012). Going forward, predators in planktonic systems must be identified and enumerated. Detailed diet studies (Llopiz et al., 2010; Llopiz, 2013) and genetic techniques (Cleary et al., 2012; Fox et al., 2012) provide a path forward and reveal new insights into predator–prey dynamics. Similarly, optical and acoustic techniques provide a means to study the distribution of plankton predators and their overlap with early life stages (Hallfredsson and Pedersen, 2009; Greer et al., 2013). Predation rates and functional forms of prey consumed relative to prey density must also be measured and parameterized for use in models (Purcell and Grover, 1990; Gentleman and Neuheimer, 2008). Finally, early life stage predation needs to be incorporated into fisheries assessments models. The work of Collie et al. (2013) and Richardson et al. (2011) indicates that early life stage predation (H_2d) can result in multiple equilibrium points in fishery populations, which create very different population dynamics than implied by the standard assessment models (H₃; see also Steele and Henderson, 1984; Bakun, 2006).

Migration hypothesis (H₁)

Despite Hjort's (1914) rejection of the migration hypothesis (H₁), there is growing evidence that changes in distribution and migration patterns affect local abundance and potentially overall abundance. Many studies have demonstrated changes in species distribution related to changing environmental conditions (Perry et al., 2005; Mueter and Litzow, 2008; Nye et al., 2009; Pinsky et al., 2013). Changes in the distribution of fisheries have also been observed (Overholtz et al., 2011; Pinsky and Fogarty, 2012). The conceptual model is that as the environment changes, the distribution of the niche changes and, as a result, species distribution changes (see Guisan and Thriller, 2005). The niche is defined as a multidimensional space within which an organism can survive and reproduce (Hutchinson, 1957); the dimensions are components of the organism's environment including biological, physical, chemical, and geological factors (Andrewartha and Birch, 1954). The basic niche is the full range of environmental conditions within which a species can survive and reproduce; the realized niche is the subset of the fundamental niche used by a species as a result of ecological interactions (e.g. competition, predation, feeding). These ideas have been used to develop niche-models of marine species and forecast changes in distribution in response to climate change (Lehodey et al., 2003; Chueng et al., 2009; Hare et al., 2012).

Atlantic mackerel in the North Atlantic provide an excellent example of changes in regional abundance owing to changes in distribution over time. In the Northwest Atlantic, Atlantic mackerel overwinter off the northeast US shelf and during summer, contingents spread to the Gulf of Maine, Gulf of St Lawrence, and near Newfoundland. Overholtz et al. (2011) documented a northward and landward shift of overwintering groups associated with warming shelf waters. This resulted in a decrease in abundance in a recreational fishery, which took place in historical overwintering grounds. Radlinski et al. (2013) emphasized changes in distribution along the northeast US shelf were partially explained by environmental changes, but fish size was also an important factor. In the Northeast Atlantic, changes in Atlantic mackerel distribution are more extreme. Astthorsson et al. (2012) reported that Atlantic mackerel were intermittently found around Iceland from 1895 to 1996. Starting in 2007, large numbers of Atlantic mackerel were found around Iceland and a commercial fishery developed. Astthorsson et al. (2012) linked the increase in Atlantic mackerel to warming which started in the 1990s but also suggested other factors may be involved including changes in population size and the age/size structure of the population. Jansen and Gislason (2013) proposed that strong year classes of mackerel stray into other areas for spawning as adults and thus, movement, distribution, age-structure, and productivity are related.

The mechanisms and consequences of changes in distribution must be pursued to more fully understand fluctuations in fishery populations. In the instances of species shifts in response to environmental change, not all species shift similarly. For example, on the northeast US shelf, 17 stocks shifted poleward, 4 stocks shifted equatorward, and 15 stocks showed no change over a 40+ year record (Nye et al., 2009). Understanding the differences among species will help elucidate the mechanisms that shape species distributions and movements. Species distributions also change as a result of changes in age-structure, size-structure, and condition; thus, changes in growth (prey processes) and survival (predation processes) are linked to changes in distribution (Bailey et al., 1982; Slotte, 1999; Radlinski et al., 2013). The energetics of migration in the context of other energetic demands (e.g. reproduction, growth, basal metabolism) also needs to be further elucidated (reproduction and ecophysiology processes) (Nisbet et al., 2012). Finally, the distribution of and movement among different habitats (e.g. spawning, overwintering, feeding) needs to be examined (e.g. Anderson et al., 2013). Similar to the predation hypothesis (H_2d) and the parental condition hypothesis (H_2a), the movement hypothesis combines multiple processes related to variability in the abundance of fishery species.

A major question for fisheries assessment and management is whether individuals are changing their distribution or the productivity of individuals in different locations is changing. The former means changing availability to fisheries and changing stock boundaries (Link et al., 2011); the latter means changing population dynamics and management reference points (Quinn and Collie, 2005). As an example of shifts in distribution, changes in bluefin tuna migration pathways have been linked to environmental, trophic, and fishing pressures and have resulted in changes in local abundance (Ravier and Fromentin, 2004; Fromentin, 2009). As an example of changes in productivity, shifts in Atlantic surfclam distributions resulted from mortality in shallower waters from thermal stress (Weinberg, 2005). Similarly, anchovy have expanded in the North Sea owing to increased productivity of existing remnant populations, which resulted from an expansion of thermal habitat (Petitgas et al., 2012). Changes in distributions can result in social and economic impacts; examples include Atlantic mackerel off Iceland (Astthorsson et al., 2012) and fishery species moving into the Gulf of Maine (Mills et al., 2013). Examining the migration hypothesis and understanding species distributions in a variable and changing environment is critical to accurately assess and successfully manage fisheries.

Ecophysiology (H_2e)

Ecophysiology is the interrelationship between the environment and an organism's physiology (Fry, 1971; Neill et al., 1994). Using temperature as an example, a range of biological processes are related to temperature either directly or indirectly (Brett, 1979; Pörtner et al., 2001; Pörtner, 2002; Brown et al., 2004). All organisms have thermal limits above and below which death is rapid. Within these limits, temperature controls a number of rate processes, including gene expression, enzyme kinetics, metabolism, activity, consumption, and growth. Organisms also respond behaviourally to temperature through migration, foraging, and resting. Temperature is also related to individual survival and fitness, as well as population growth rate (Brown et al., 2004; Kingsolver, 2009). Temperature is a component of the basic niche, which is important in shaping species distribution and abundance (Hutchinson, 1957; Kerr and Werner, 1980; Magnuson and DeStasio, 1997). Similar biological processes are related to other environmental factors, including dissolved oxygen (Craig and Crowder, 2005; Prince and Goodyear, 2006), salinity (Attrill and Rundle, 2002; Lowe et al., 2012), and contaminants (Matthiessen et al., 2002; Brooks et al., 2012). This view of ecophysiology is closely related to the metabolic theory of ecology (Brown et al., 2004).

One example of temperature affecting abundance and distribution is overwinter mortality in temperate ecosystems. Overwinter mortality is caused by the physiological response to temperature combined with the relative scarcity of food and potential susceptibility to predation (Hurst, 2007). Along the east coast of the United States, overwinter mortality affects the distribution and abundance a number of fishery species including pink shrimp (Hettler, 1992), reef fish (McBride and Able, 1998; Parker and Dixon, 1998), striped bass (Hurst and Conover, 1998), Atlantic croaker (Hare and Able, 2007), blue crab (Bauer and Miller, 2010), red drum (Anderson and Scharf, 2014), and grey snapper (Wuenschel et al., 2012). For Atlantic croaker, laboratory studies found decreased juvenile survival below 3°C (Lankford and Targett, 2001a, b). Field studies in spring confirm that juvenile abundance is related to winter water temperature (Hare and Able, 2007). As a result, recruitment to the adult population is also related to winter temperature and a population model including an empirical temperature-dependent recruitment function replicated the past patterns in Atlantic croaker abundance (Figure 5, Hare et al., 2010). Further, Hare et al. (2010) developed a one-dimensional distribution model using population size and winter temperature as independent variables. The example of overwinter mortality indicates that the environment can limit fishery population abundance and distribution through ecophysiological processes.

Figure 5.

The relationship between winter temperature and recruitment of Atlantic croaker. The correlation supports the hypothesis that overwinter survival determines year-class success in Atlantic croaker. A recruitment-spawning-stock biomass function including winter temperature fits observed recruitment much better than a function without the environmental term. The colour of the symbols represents winter temperature. The lines in the recruitment-spawning-stock biomass panel shows estimated recruitment at a mean winter temperatures of −4°C, 0°C, and 4°C (modified from Hare et al., 2010). The mechanistic steps in the hypothesis from larval ingress into juvenile habitats to recruitment are indicated by the boxes on the right side.

Ecophysiology has been part of fisheries oceanography since Hjort (1914, 1926), but the emphasis is on the physiological interaction between the environment and the organism (ecophysiology), rather than the organism obtaining resources from the environment (e.g. starvation and the critical period hypothesis). Since these ideas are not new, there have been many calls for increasing our understanding of ecophysiology (Neill et al., 1994; Helmuth et al., 2006; Helmuth, 2009; Pörtner, 2010). That said, there are several research areas that would lead to better integration between ecophysiology and the other processes of importance to fishery abundance and distribution. First, the allocation of energy among basal metabolism, growth, reproduction, daily activity, and seasonal activity needs to be understood and quantified. A similar research need was identified for the parental condition hypothesis and the migration hypothesis. Second, the effect of temperature and other environmental variables on physiology and energy allocation needs to be examined. Third, the combined role of prey, predators, physical environment, and physiology needs to be developed in terms of individual survival and population growth rates. This information could be used to parameterized process-oriented models of distribution, growth, survival, and recruitment (e.g. Okunishi et al., 2012) in contrast to populations models using empirical environment–biological response functions (e.g. temperature-recruitment, Hare et al., 2010).

Conclusions: the legacy of Johan Hjort

The goal of fisheries oceanography is to understand the oceanographic and ecological processes that affect fishery abundance, distribution, and availability and then apply this understanding to improve fisheries assessment and management. In this essay, I considered fisheries oceanography from the perspective of fishery abundance and distribution. The examination of the four hypotheses above (H_2a, H_2d, H₁, H_2e) indicates that abundance and distribution are related. This idea is well captured by the Basin Hypothesis (MacCall, 1990). Taking some liberties with this hypothesis, distribution is linked to the abundance of the population and dimensions of the niche (or the distribution of essential fish habitat). This concept was also recognized by Hjort who discussed the migration hypothesis in terms of “sensitivities” of species responding to changes in the environment. The niche concept is defined above; essential fish habitat is defined as all the areas necessary for fish reproduction, feeding, growth, and survival (Beck et al., 2001; Manderson et al., 2002). Thus, essential fish habitat represents a subset of the basic niche that is similar to the realized niche. The distribution of the niche and essential fish habitat is related to distribution and abundance of fishery species (MacCall, 1990; Bartolino et al., 2011) and there is strong evidence that overall population abundance is controlled by the area available, or in other words the multidimensional area of the niche and the amount of habitat (Sinclair, 1988; MacKenzie et al., 2003).

From the review of the four hypothesis above (H_2a, H_2d, H₁, and H_2e), it is also clear that the regulating processes of prey, predators, movement, ecophysiology, and reproduction vary through time and space. This implies that the niche and essential fish habitat concepts must be viewed as dynamic: varying and changing as a result of the interaction among the regulating processes (Figure 2, MacKenzie et al., 2007; Cheung et al., 2011). Further, these regulating processes act in most life stages and can act in one life stage and have consequences for subsequent life stages. In all likelihood, in different portions of a species range, the relative contribution of different processes at different life stages varies (e.g. Neill et al., 1994; Myers, 1998). Similarly, as climate and fishing pressure vary and change, the temporal contribution of the different processes acting at different life stages likely changes (e.g. Petitgas et al., 2012). This results in a complex pattern of fishery abundance and distribution that varies and changes owing to variability and changes in the regulating processes and subsequently the dimensions of the niche and the quality and quantity of habitat (Figure 6). To improve forecasts and predictions of abundance and distribution, we need to understand the action and interaction of the five regulating processes. Further, we need to incorporate these processes and their related hypotheses in fisheries assessment and management. We can simplify by focusing on one or a subset of hypotheses and processes, but we must be careful to recognize the limitations of our simplifications, and improve or change our approach when science dictates (Ianelli et al., 2011; Link et al., 2011).

Figure 6.

A schematic of the Basin Hypothesis (MacCall, 1990) showing changes in the shape and the distribution of the “basin” through the life history or for one life stage through time. The red line shows the distribution at low population abundance. The green line shows the distribution at high population abundance. The “basin” represents habitat quality or population growth rate and is analogous to the niche of the species. The “basin” includes the effect of all the processes affecting abundance and distribution (prey, predation, movement, ecophysiology, and reproduction). The shape of the “basin” changes through time which will affect the abundance and distribution of fishery species. This schematic is further complicated by the complex life history of most marine organisms; each life stage potentially having its own basin.

Much of the focus here has been on populations and species as they relate to single-species assessment and management. There has been a broad call for Ecosystem-Based Fisheries Management (EPAP, 1999; Pikitch et al., 2004) and a call for a transition from fisheries oceanography to ecosystem oceanography (Cury et al., 2008). The ideas discussed here are relevant to broader ecosystem approaches. The “basin” described above (Figure 6) is for a single species or population but includes interactions with components of the ecosystem and the environment. A multispecies approach can be envisioned as multiple interacting “basins”, one for each population or species. An ecosystem-approach can be envisioned as interacting “basins” for functional groups or communities. An important issue regarding multispecies, functional group, or community level aggregation is the ability to combine populations/species and life stages and treat them similarly from an ecological, fishery, and socio-economic perspective. This combination is similar to the difference between a stock (a management unit) and a population (an ecological unit) (see Secor, 2013), but at higher-levels of organization. Understanding the implications of such aggregation, investigating different approaches to aggregation, and parameterizing aggregation should be a focus of future fisheries oceanography research and an integral part of the development of Ecosystem-Based Fisheries Management.

Another focus moving forward needs to be the development of models that integrate understanding and improve assessments, be they single species, multispecies, or ecosystem assessments. These models need to span a range of complexity and address a range of hypotheses (Figure 2). Most of the models currently used in fishery stock assessment are based on fishing mortality and a spawning-stock biomass–recruitment relationship (H₃), but recently other hypotheses and processes have been integrated into the stock assessment framework (e.g. Murawski et al., 2001; Quinn and Collie, 2005; Ianelli et al., 2011). Most oceanographic and climate models include bottom-up effects on early life stages (H_2b) or environmentally driven niche-models (H_2e). Much of the ecological and ecosystem modelling includes foodweb dynamics (H₄) but excludes early life stages. The structure of a model will limit the potential findings (aka structural uncertainty; see McAllister and Kirchner, 2002; Essington and Plagányi, 2014). For example, the structure of nutrient–phytoplankton–zooplankton models determines whether bottom-up or top-down processes are found to be important (Sailley et al., 2013). Given the variety of hypotheses for the determinants of abundance and distribution of fishery species, the amount of structural uncertainty in our present assessments and management is higher than we typically recognize. This uncertainty needs to at least be recognized and then future efforts should work to decrease this uncertainty.

One could argue that the research questions and path forward identified here represents “more of the same”: more reductionist science and building more complicated models. This is a valid criticism, but I would counter that the emphasis moving forward needs to be integrating among hypotheses and processes incorporated into models. Yes, we need to continue our investigation of individual hypotheses (reductionist science), but at the same time, we need to broaden our view and integrate among hypotheses and processes. Additionally, we need to pursue hypotheses and not automatically keep working within the traditional hypotheses. Thus, I recommend we focus on Hjort's approach, as well as the hypotheses attributed to him. Hjort proposed and pursued multiple hypotheses using interdisciplinary science. He moved away from the migration hypothesis (H₁) based on evidence at the time and was willing to propose and investigate different hypotheses. Perhaps he knew of Chamberlin's (1890) work on “multiple working hypotheses”, an idea which is again gaining momentum (Burnham and Anderson, 2002). Following Hjort, a hundred years of reductionist science investigating single hypotheses and processes has created isolated paradigms. We need to follow Hjort's approach and combine and integrate our understanding of the various processes that affect fishery abundance and distribution. We then need to apply to this science to improve fishery assessment and management.

Acknowledgements

This essay benefited for many people over the years who have shaped my thinking on fisheries oceanography. In particular, Bob Cowen, Ken Able, Jeremy Collie, Mike Fahay, Mike Fogarty, Kevin Friedland, Jeff Govoni, Anne Hollowed, John Manderson, Dave Richardson, Cisco Werner, and Mark Wuenschel. Also the participants of the NOAA NMFS FATE programme (Fisheries and the Environment) provide stimulating science and ideas helping to move fisheries oceanography forward; in particular, I would like to acknowledge the FATE programme managers Kenric Osgood and Michael Ford. I also thank three anonymous reviewers, David Richardson, John Manderson, Rich Bell, and Howard Browman for their constructive and thought-provoking comments on earlier drafts; this work greatly benefited from their input. Acknowledgement of the above individuals does not imply their endorsement of this work; I have sole responsibility for this contribution. The views expressed herein are mine and do not necessarily reflect the views of NOAA or any of its subagencies.

Author notes

Handling editor: Howard Browman