Reviews in Fish Biology and Fisheries

, Volume 19, Issue 1, pp 69–95

Connectivity, sustainability, and yield: bridging the gap between conventional fisheries management and marine protected areas

Authors

    • Department of Wildlife, Fish and Conservation BiologyUniversity of California
  • Daniel R. Brumbaugh
    • Center for Biodiversity and ConservationAmerican Museum of Natural History
  • Churchill Grimes
    • National Oceanic and Atmospheric Administration, Fisheries Ecology DivisionSouthwest Fisheries Science Center
  • Julie B. Kellner
    • Department of Environmental Science and PolicyUniversity of California
  • John Largier
    • Bodega Marine LaboratoryUniversity of California, Davis
  • Michael R. O’Farrell
    • National Oceanic and Atmospheric Administration, Fisheries Ecology DivisionSouthwest Fisheries Science Center
  • Stephen Ralston
    • National Oceanic and Atmospheric Administration, Fisheries Ecology DivisionSouthwest Fisheries Science Center
  • Elaine Soulanille
    • National Oceanic and Atmospheric Administration, Fisheries Ecology DivisionSouthwest Fisheries Science Center
  • Vidar Wespestad
    • Resource Analysts International
Article

DOI: 10.1007/s11160-008-9092-z

Cite this article as:
Botsford, L.W., Brumbaugh, D.R., Grimes, C. et al. Rev Fish Biol Fisheries (2009) 19: 69. doi:10.1007/s11160-008-9092-z

Abstract

A substantial shift toward use of marine protected areas (MPAs) for conservation and fisheries management is currently underway. This shift to explicit spatial management presents new challenges and uncertainties for ecologists and resource managers. In particular, the potential for MPAs to change population sustainability, fishery yield, and ecosystem properties depends on the poorly understood consequences of three critical forms of connectivity over space: larval dispersal, juvenile and adult swimming, and movement of fishermen. Conventional fishery management describes the dynamics and current status of fish populations, with increasing recent emphasis on sustainability, often through reference points that reflect individual replacement. These compare lifetime egg production (LEP) to a critical replacement threshold (CRT) whose value is uncertain. Sustainability of spatially distributed populations also depends on individual replacement, but through all possible paths created by larval dispersal and LEP at each location. Model calculations of spatial replacement considering larval connectivity alone indicate sustainability and yield depend on species dispersal distance and the distribution of LEP created by species habitat distribution and fishing mortality. Adding MPAs creates areas with high LEP, increasing sustainability, but not necessarily yield. Generally, short distance dispersers will persist in almost all MPAs, while sustainability of long distance dispersers requires a specific density of MPAs along the coast. The value of that density also depends on the uncertain CRT, as well as fishing rate. MPAs can increase yield in areas with previously low LEP but for short distance dispersers, high yields will require many small MPAs. The paucity of information on larval dispersal distances, especially in cases with strong advection, renders these projections uncertain. Adding juvenile and adult movement to these calculations reduces LEP near the edges in MPAs, if movement is within a home-range, but more broadly over space if movement is diffusive. Adding movement of fishermen shifts effort on the basis of anticipated revenues and fishing costs, leading to lower LEP near ports, for example. Our evolving understanding of connectivity in spatial management could form the basis for a new, spatially oriented replacement reference point for sustainability, with associated new uncertainties.

Keywords

ConnectivityMarine reserveMetapopulationFisheries managementDispersalHome range

Introduction

Management of marine resources for biodiversity objectives and fishery production is in a period of dynamic change. A growing awareness of the number of managed fisheries that have declined to low levels has led to disenchantment with conventional fishery management and calls for greater use of areas with marine protected areas (MPAs)—areas where no fishing is allowed—as a tool for better conservation and more reliable fishery yields. For example, Worm et al. (2006) recently made the controversial claim that fisheries are declining fast enough to be exhausted by the year 2048, and that the remediation for this failure of conventional management was MPAs (see critiques in Science vol. 316, No. 5829 and Longhurst 2007). Additionally, the 2002 World Summit on Sustainable Development called for increased implementation of MPAs. MPAs are widely perceived to be effective in preventing population declines and are mostly unencumbered by a past history that includes failures. However, an empirical meta-analysis of 70 MPAs indicated that 37% produced no increase in density, 4–10% produced no increase in biomass, 11–20% produced no increase in mean size, and 24–41% produced no increase in diversity (Halpern 2003). There is a need to understand the dynamics and biophysical components of the more heterogeneous spatial environment created by MPAs, and why they work better in some instances than others.

Management with MPAs by definition involves fishing rates that vary over space, thus increasing the importance of spatial aspects of the spatio-temporal dynamics of marine populations. Most marine populations are spatially distributed metapopulations (Kritzer and Sale 2004), but their dynamics have been approximated by treating them as single, well-mixed populations (or identical, independent subpopulations). However, this approach becomes problematic when fishing mortality rate (F) varies over space. Even management consequences calculated from spatially-explicit simulations are of limited value because of our incomplete understanding of the general characteristics of the spatial dynamics of marine populations, and the uncertainty in relevant rates of movement over space.

The naïve hope for a proposed MPA is that if one reduces fishing to zero in a small part of an ecosystem that area will return to the original pristine ecosystem. However, this is clearly not the case. The dynamic behavior of spatially distributed marine populations is not merely the sum of a number of adjacent, independent single populations. Their dynamics are more complex than that because of the movement of individuals over space. That movement provides a web of connectivity among locations within the metapopulation, which leads to complex dynamic dependencies across reserve boundaries. Here we focus on different components of that connectivity, providing a summary of our developing understanding of them, and their consequences for population dynamics and management.

Three kinds of movement have a significant effect on the dynamics of exploited spatially distributed marine populations. Early in life most marine species have a larval stage that disperses over distances that can be substantial, thus linking the dynamics of populations on a scale of that dispersal distance. Following metamorphosis to the post-larval form, fish become substantially more motile in many, but not all, cases providing additional connectivity over a broad range of spatial scales, from local daily movement over meters to migrations over thousands of kilometers. The third type of movement of dynamic importance in exploited, spatially distributed populations is the movement of fishermen.

These three types of connectivity differ substantially in temporal scale and driving mechanisms so we have examined them separately, and refer to them as larval connectivity, juvenile and adult connectivity, and economic connectivity. We assess their consequences primarily in terms of the goals and currencies of greatest interest in fisheries and conservation of marine resources: sustainability and yield. Sustainability is the ability of the population and the associated fishery to persist into the long-term future. Yield is a benefit in the management of a fishery, and a decline in yield is a potential cost when setting aside an MPA solely for conservation. We begin by reviewing the ways in which conventional fisheries management seeks sustainability and high yield, and the uncertainties involved. We then describe the dynamics of sustainability in spatially distributed populations with MPAs as we add the three types of connectivity. Finally, we present synthetic conclusions that encapsulate our current understanding of connectivity and MPAs, and identify future challenges in this arena.

Conventional fisheries management

Stock assessment and harvest policy

In simple terms, the goals of conventional fisheries stock assessment traditionally have been (1) to estimate the relationship between the rate of fishing and sustainable yield, and (2) to determine the current status of an exploited fish population within the context of that relationship. If the relationship between F and sustainable yield were known with great precision, it would be a simple matter to optimize yield by fishing at the rate that produces the maximum sustainable yield (MSY). However, over the past 30 years there has been increasing appreciation of the effects of substantial uncertainty on fishery stock assessment and management. In many cases the relationship between sustainable yield and fishing is not known with sufficient confidence to prevent pressure for increased yields from driving fishing rates to excessive levels (e.g., Ludwig et al. 1993; Botsford et al. 1997). There has been a global effort to combat these effects [e.g., in the U.S., the Sustainable Fisheries Act (SFA) of 1996; internationally, the Code of Conduct for Responsible Fisheries (FAO 1995)]. Target and limit reference points for the rate of fishing and population depletion are employed globally, and are now required for all federally managed fisheries in the United States. Target reference points are similar to the conventional goals of fisheries such as MSY. Limit reference points represent a threshold condition to be avoided, beyond which the fishery is considered to be in jeopardy. Limit reference points can be conditions on the rate of removal by fishing or the abundance to which the fished population is allowed to decline. The precautionary approach to fishery management as codified in the Code of Conduct for Responsible Fisheries (FAO 1995), notes that (1) failures in fishery management systems are most often due to insufficient precaution in the face of high levels of uncertainty and (2) when fishery impacts are uncertain, priority should be given to conserving the productive capacity of the resource.

Increasing appreciation of uncertainty in fisheries management has led to a more explicit focus on sustainability (e.g., through target and limit reference points), as well as improvements in the way sustainability is determined quantitatively. Since the 1980s, the focus has sharpened on the critical aspect of the stock-recruitment relationship required for population persistence: behavior at low abundance. Shepherd (1982) showed that one could combine the results of a simple life table calculation, i.e., lifetime reproduction, with empirical estimates of recruitment and spawning from a stock assessment, to infer population persistence under different rates of fishing. Lifetime reproduction, a measure known to be important in the persistence of linear (non-density-dependent) populations (i.e., as R0, Caswell 2002), was written in units of eggs rather than individuals reproduced, and termed eggs-per-recruit (EPR; here we refer to this quantity in terms of its population context as Lifetime Egg Production, see Table 1). It enters the graphical solution for the equilibrium level of recruitment in a population with density-dependent recruitment through its effect on the slope of the “replacement line.” The replacement line is a line through the origin of the spawner-recruit curve with slope 1/LEP. As fishing mortality increases, LEP decreases, and the slope of the replacement line increases (Fig. 1). The population equilibrium corresponding to that level of fishing mortality occurs where the replacement line intersects the underlying spawner-recruit curve. If the slope of the replacement line were to exceed the slope of the spawner-recruit curve at the origin, the indicated equilibrium would be zero, i.e., if the value of LEP declines below that critical replacement threshold (CRT) the population will collapse.
Table 1

Acronyms and definitions of terms commonly used in both conventional fisheries management and management with reserves

Term

Abbreviation

Definition

Eggs per Recruita

EPR

The number of eggs produced by a recruit over the course of its lifetime.

Fractional Lifetime Egg Productionb

FLEP

The ratio of LEP of a fished recruit to the unfished level of LEP. In practice, since the unfished LEP may be difficult or impossible to estimate, FLEP can be specified as the ratio of LEP of a fished recruit to the LEP estimated at an earlier time, prior to heavy exploitation.

Lifetime Egg Productiona

LEP

The number of eggs produced by a recruit over the course of its lifetime. Equivalent to EPR and used in the calculation of FLEP.

Lifetime Reproductiona

R0

The number of female offspring produced by a female over her lifetime. This term is most commonly used in age structured models outside the realm of fisheries.

Spawning per Recruitb

SPR

The ratio of fished SSBR (or alternatively EPR) to unfished SSBR. Also referred to as the Spawning Potential Ratio.

Spawning Stock Biomass per Recruita

SSBR

The level of spawning stock biomass produced by a recruit over the course of its lifetime. Is often used as part of the calculation of SPR.

aRaw lifetime reproduction value

bNormalized lifetime reproduction value

https://static-content.springer.com/image/art%3A10.1007%2Fs11160-008-9092-z/MediaObjects/11160_2008_9092_Fig1_HTML.gif
Fig. 1

The relationship between recruitment and the spawning stock for a hypothetical fish population. Symbols represent “noisy” stock-recruit estimates from an assessment, with the solid gray line depicting the true underlying spawner-recruit relationship. Three replacement lines are plotted that show how equilibria (the intersection of the replacement line and the spawner-recruit curve) change as the fishing mortality rate (F) is increased and lifetime egg production (LEP) is diminished (the slope of the replacement lines is 1/LEP). F100% is the replacement line for an unfished stock and the asterisk is the unexploited equilibrium. Frep is the replacement line which bisects the observed data points. F40% is the replacement line that represents the fishing mortality rate that reduces LEP to 40% of its unfished value. FCRT is the replacement line associated with the critical replacement threshold. Note that when the slope of the replacement line becomes greater than the slope of the spawner-recruit line at the origin, the equilibium will be zero, indicating LEP is no longer large enough for individuals to replace themselves

An empirical basis for the CRT has been sought through two avenues: (1) an estimate specific to the species of interest and (2) a general estimate applicable to species with few data, expressed as the fraction of natural, unfished LEP. Species-specific estimates can be estimated from a “steepness coefficient” describing the behavior of the spawner-recruit curve near the origin (see Dorn 2002 for a description of steepness). To generalize the existing specific estimates, in the hope that they bear implications for other species, the CRT is commonly expressed in terms of the fraction of the natural, unfished LEP, i.e., FLEP, which is the same as spawning potential ratio in the fisheries literature (Table 1). This allows fishery scientists and managers to express CRTs in common terms for all fisheries, and ask what suitable, safe values might be. Mace and Sissenwine (1993) presented information from 91 spawner-recruit data sets, which showed that an estimated CRT in terms of SPR averaged approximately 20%. They proposed that a strategy of using F30% (the fishing mortality that reduces FLEP to 30% of the unfished value—note that F30% < F20%) would be a more conservative limit. Values as high as 40% have been recommended based on similar analyses (Clark 1991, 1993; Mace 1994). As a consequence of these and other studies of stock-recruitment data, and impact on harvest, F40% has frequently been adopted as a “proxy” target exploitation rate by fishery managers in situations where estimation of stock-specific productivity parameters is believed to be unreliable. However, there are instances where this approach has failed, most notably for species of rockfish (Sebastes spp.) on the U.S. west coast, where experience has shown that lower exploitation rates may be needed to prevent stock collapse (Dorn 2002; Ralston 2002).

Incorporating space in conventional management

With respect to the spatial distribution of fishing and fish, the preponderance of stock assessments assumes that fish populations are spatially homogeneous. It is not uncommon for assessment models to ignore the effects of spatial variability in demographic rates, including recruitment, growth, maturity, natural mortality, and fishing mortality. While spatial heterogeneity is acknowledged, as a practical matter, most models simplify this complexity (see Hilborn and Walters 1992). Population modeling can be conducted on a stock-wide basis, with estimation of a single set of population and management parameters, but with “geographic apportionment” used to partition the total stock into smaller spatial units based on regional fishery and/or survey information (Quinn and Deriso 1999). Fundamentally, the reason that most contemporary assessment models lack even a moderate degree of spatial sophistication is the fact that there are inadequate data available to support that kind of modeling effort; not because analysts do not recognize or appreciate the existence of important biological processes operating at finer spatial scales (Walters and Martell 2004).

Use of closed areas

A number of different approaches to spatial fisheries management have been tried over the years with varying degrees of success. One spatial approach to management has been to reduce adverse impacts to habitat caused by destructive fishing practices, as is required under the essential fish habitat provisions of the U.S. Sustainable Fisheries Act. Those provisions also call for protection of habitat areas of particular concern from damage due to fishing gear (e.g., deep-sea coral and other biogenic habitats), irrespective of the importance of those sites to fish production. A second type of spatial fishery management is closure of specific locations where animals congregate and are particularly susceptible to fishing impacts, including aggregations in spawning or nursery habitats.

Spatial closures are also used to address the problem of “technical interaction,” the situation in which shared vulnerability to a fishing gear causes a set of species to be caught together as an assemblage (Pope 1979; Pelletier and Mahévas 2005). A difficulty arises when some species in the assemblage are resilient to the effects of exploitation and others are not, requiring restrictions on fishing and usually referred to as weak stock management. For example, because of concerns about bycatch of a number of overfished stocks [e.g., canary rockfish (Sebastes pinniger), bocaccio (S. paucispinis), darkblotched rockfish (S. crameri)] the outer shelf and upper continental slope of the west coast of the U.S. from Mexico to Canada have been closed to fishing for several years. The intent of the closure is to minimize gear contacts with overfished rockfish stocks, even though this has reduced harvesting opportunities for abundant species that could support healthy sustainable fisheries [e.g., yellowtail rockfish (S. flavidus) and chilipepper (S. goodei)].

Consequences of connectivity for sustainability and yield

With a sense of how questions regarding sustainability and yield are currently addressed by conventional fisheries management, we now turn to the question of how connectivity is likely to influence sustainability and yield in spatial management. The data available from existing spatial management consist primarily of evaluation of populations within the protected areas relative to those outside (Halpern 2003; Micheli et al. 2004). As noted in the introduction, these data are sufficient to identify how frequently MPAs fail to increase biomass, abundance, mean size and diversity, however they are insufficient to determine why. In particular, they do not indicate the role played by differences in connectivity. We therefore need to depend on mathematical models of spatially distributed populations with spatially varying fishing and various kinds of movement. Our aim is to define the dependence of yield and sustainability on the different kinds of movement, then to compare the results to what is known about the three types of movement.

Effects of larval dispersal on sustainability

To understand the consequences of movement in the larval phase, we first assume that we are dealing with a species with sedentary adults and juveniles, and that fishing effort is removed with the implementation of MPAs (i.e., effort is not shifted to other locations). We will revisit these assumptions in later sections.

Several theoretical analyses of the general conditions necessary for persistence of metapopulations indicate that these conditions can be interpreted as extending the replacement concept to account for the fact that individuals in a population can replace themselves through many paths (Botsford et al. 2001; Hastings and Botsford 2006). In the non-spatial, single population case, larvae produced return to the population or die, whereas in the explicitly spatial metapopulation larvae disperse, with some of the larvae produced settling in reserves while others settle in fished areas. The dispersing larvae contribute to replacement by surviving at the settled location and reproducing to produce larvae according to the local LEP. Some of these will return to the origin of their parents. It is important to realize that population persistence depends on all such multi-generational closed loops, the return paths, as well as the source paths (Hastings and Botsford 2006).

The expressions that have been derived to calculate this replacement condition depend only on the larval dispersal pattern, the values of LEP at each location, and the CRT (Botsford et al. 2001; Hastings and Botsford 2006). The fact that the CRT appears in these formulations is important as the same dominant uncertainty in conventional management also appears in spatial management.

The results of these spatial persistence calculations for a system of MPAs are illustrated in an assessment method termed dispersal-per-recruit (DPR; Kaplan et al. 2006; Fig. 2). This application uses specific information on the width and spacing of reserves, the spatial distribution of habitat, and estimated or assumed levels of fishing outside of reserves to calculate where populations with specific mean dispersal distances will be likely to persist. An example of species inhabiting a specific depth range along the California coast illustrates how reserve designs and habitat patterns (Fig. 2a) can be represented in one-dimension to calculate the distribution of persistent populations (Fig. 2b). The relative distributions of MPAs (red) and habitat (green) lead to a spatial distribution of FLEP at each point in space (shaded areas in Fig. 2c) that takes one of three values depending on whether the point occurs outside suitable habitat (FLEP = 0.0) in suitable habitat and inside a reserve (FLEP = 1.0) or in suitable habitat but outside reserves (FLEP = the value resulting from the fishery, here assumed to be 0.2). The assessment method computes the level of replacement and associated settlement at each point that would result for species with different dispersal distances. The two general characteristics observed in these calculations are illustrated in Fig. 2c: (1) short distance dispersers tend to persist in all reserves, while (2) long distance dispersers tend to persist only in larger reserves or where a certain fraction of the local coastline is in reserves (also see Kaplan et al. 2008). These follow from earlier general results that: (1) single MPAs will allow species with dispersal distance up to and including the width of the MPA to persist, and (2) species with any dispersal distance will persist once the fraction of coastline in reserves reaches the critical replacement fraction (e.g., 40%), or less if fishing is not extremely high (Botsford et al. 2001). Because the latter response (2) involves the effects of many reserves acting together, it was termed a network effect. Both Fig. 2 and the earlier general results used a decaying exponential (i.e., a Laplace distribution) to represent dispersal, approximating pure diffusion without advection. The shape of the dispersal distribution makes little difference to persistence (Lockwood et al. 2002). However, when alongshore advection is present, persistence can deteriorate rapidly (Gaylord and Gaines 2000; Botsford et al. 2001; Kaplan 2006). To test a proposal that reserve spacing would need to be varied to accommodate the many dispersal distances of species in natural systems (Palumbi 2003), Kaplan and Botsford (2005) showed that randomly spaced reserves produced the same results as evenly spaced reserves, except in the fortuitous occurrence of a part of the coastline having MPA coverage locally greater than the critical replacement fraction, when the global mean was less than that critical replacement fraction.
https://static-content.springer.com/image/art%3A10.1007%2Fs11160-008-9092-z/MediaObjects/11160_2008_9092_Fig2_HTML.gif
Fig. 2

An example of spatial population sustainability of a proposed reserve design along the California coast. The two-dimensional reserve (red) and rock habitat from 0 to 30 m depth (green) configurations (a) are approximated in one dimension (b). The reserve and habitat configurations produce a pattern of Fraction of Lifetime Egg Production (FLEP; shaded areas in c and d) corresponding to areas with no habitat (FLEP = 0), fished habitat (FLEP = 0.35), and fully protected (FLEP = 1.0). The colored lines in (c) are the equilibrium fraction of natural settlement at each point for species with mean dispersal distances of 25 km (green) and 1 km (blue), computed using the Dispersal per Recruit (DPR) method (Kaplan et al. 2006) with a hockey-stick settlement-recruit function and a critical replacement threshold of 35%. The settlement habitat is saturated where settlement is greater than 35%. Panel (d) demonstrates equilibrium yield at each point in space for the same two dispersal distances as in (c)

Effects of larval dispersal on yield

The question of how MPAs affect yield was initially addressed using a larval pool model in which all larvae enter a single pool after which they are evenly distributed along a finite coastline. This approximately represents the behavior of species with long dispersal distances. The effects on yield of setting aside part of a fished area as a MPA can be qualitatively described in terms of an approximate equivalence in yields possible from management with MPAs and conventional fishery management. This link between yield from reserves and conventional management was demonstrated to be exact equivalence for a simple age-structured model (Hastings and Botsford 1999). The equivalence notion is similar to a result demonstrating a fundamental invariant, the product of fishing mortality rate and area not in reserves, in a logistic model with both reserves and conventional control of fishing mortality rate (Mangel 1998, 2000).

An important corollary of this relationship is that MPAs will provide little or no improvement to well managed fisheries with low fishing mortality rate, but rather will benefit only fisheries that are heavily fished or overfished (Botsford et al. 2003). This result has appeared in a number of simulations of MPAs both with the larval pool assumption (e.g., Holland and Brazee 1996), and with more realistic representations of larval dispersal (e.g., Quinn et al. 1993). The value of fishing mortality rate above which reserves will provide a positive contribution to yield has recently been derived for a size-structured model with density-dependent recruitment and larval pool dispersal (Hart 2006).

This approximate equivalence is valuable as a rule of thumb for yields to be expected from MPAs in data poor situations, but of course there are exceptions that shift the balance in favor of reserves or in the opposite direction in favor of conventional management. Most of these involve the effects of density dependence. Pre-dispersal density dependence tends to favor conventional fishery management because the higher densities created within reserves can reduce somatic growth and egg production (Parrish 1999; Gådmark et al. 2006). Conversely, the dependence of settlement on post-settlement adult density, instead of larval settlement density, tends to favor reserves owing to a combination of (1) reserves serving as a source of potential settlers and (2) low levels of density dependent mortality experienced by settlers in fished areas (Gaylord et al. 2005; Ralston and O’Farrell 2008; White and Kendall 2007).

The remaining connectivity question is how yield changes with departure from the assumption of larval-pool dispersal, i.e., what happens to short-distance dispersers? A specific example for an overfished population (FLEP = 0.2 < CRT = 0.35) is the distribution of yield in Fig. 2d. The distribution of yield is determined by the distribution of the product of yield per recruit (YPR) and recruitment. Here, recruitment equals settlement when settlement is less than 35%, and the maximum value when the post-settlement habitat is saturated (i.e., a hockey-stick settler-recruit relationship was used).

The effects of varying MPAs and fishing are illustrated by simulation results from a size-structured model in which the MPAs were evenly spaced at 25 spatial units, then increased in size from zero to 2, 5, then 10 spatial units covering 8, 20, then 40% of the coastline (Fig. 3; Botsford et al. 2004). The CRT was 35% for this model population, and density-dependent recruitment depended on settlement levels. For long-distance dispersers, at low fishing mortality rate, yield was higher with conventional management, but when fishing mortality rate exceeded the MSY, yield was greater with MPAs. However, for short-distance dispersers, while they became persistent when only 8% of the coastline was in MPAs, yield was never as great as it became for long distance dispersers as the area in MPAs increased. This was due to the fact that dispersal beyond the edges of the MPAs was less and extended over less area for short-distance dispersers.
https://static-content.springer.com/image/art%3A10.1007%2Fs11160-008-9092-z/MediaObjects/11160_2008_9092_Fig3_HTML.gif
Fig. 3

Catch from a simulated age-structured population with reserves spaced every 25 spatial units along a coastline. The replacement required for persistence of the simulated population is 35%. Reserve sizes are 0 spatial units (gray), 2 spatial units covering 8% of the coast (red), 5 spatial units covering 25% of the coast (green), and 10 spatial units covering 40% of the coast (blue). From Botsford et al. (2004)

The maximum yield possible with MPAs was shown analytically to depend on the effects of reserve size and spacing on both sustainability and yield. Maximum yield at high fishing mortality rates (i.e., FLEP = 0 outside MPAs) will be obtained when there is an adequate fraction of the coastline in reserves and reserves are as small as possible (Hastings and Botsford 2003; Neubert 2003). This makes sense when interpreted in terms of maximizing larval export.

This result can be compared to yields at other values of FLEP and other MPA sizes and spacing distances in an application of the DPR method to periodic reserves along an infinite coastline, with several different levels of fishing outside (Fig. 4; Kaplan et al. 2006). An approximate practical interpretation of this plot is that for each level of fishing (resulting in FLEP = 0.0, 0.1, or 0.2), as long as reserves are large enough for sustainability, increasing the fraction of coastline in MPAs will increase yield until the fraction of coastline in MPAs approximately equals the fraction required for sustainability for all reserve sizes and dispersal distances. Further increase would require having smaller reserves, approaching the results in Hastings and Botsford (2003).
https://static-content.springer.com/image/art%3A10.1007%2Fs11160-008-9092-z/MediaObjects/11160_2008_9092_Fig4_HTML.gif
Fig. 4

Yield per unit habitat length for a system of uniformly spaced, equally sized reserves as a function of the ratio of reserve width to dispersal distance and the fraction of the coastline in reserves. The grey area represents parameter space resulting in a collapsed population. Yield contours are normalized to the maximum yield value for all three panels. The area above the heavy dotted line corresponds to configurations which result in all locations being saturated with recruits. The level of fishing in non-reserve areas is represented by the degree to which the fraction of unfished lifetime egg production (FLEP) is reduced. In order of decreasing fishing mortality: (a) FLEP = 0, (b) FLEP = 0.1, (c) FLEP = 0.2. From Kaplan et al. (2006)

Yield results change dramatically when there is directional dispersal, as also seen for the analysis of sustainability. With strong advection, it is possible to obtain higher yields than would result from conventional management as long as the structure of the connectivity is known. Morgan and Botsford (2001) examined the case of a strong source and three sinks, in which placing an MPA in the source provided greater yield than identical conventional fishery management over all four patches. This indicates that greater catches are possible with reserves in the case with a strong source, but that to take advantage of that, the location of the source must be known. Gaines et al. (2003) examined yields possible through reserves along a coastline in which there was strong advection in one direction for several years, with advection in the reverse direction in other years. Placing a reserve at each end of this coastline and one in the middle of the coastline resulted in yields greater than those obtainable by conventional management. This result may be due to the source/sink structure created by the combinations of years with strong advection.

Comparison of effects of larval dispersal with available dispersal data

The characteristics of larval dispersal that affect sustainability and yield in fisheries can be compared with our knowledge of larval dispersal to determine likely behavior of fisheries with spatial management as well as the associated uncertainties. Ideally, to project future sustainability and yield from MPAs, we need to know whether dispersal is predominantly diffusion without substantial advection (or not so), and what the associated spatial scale of diffusion is. We need to know if strong advective processes are present, their interannual variability, and whether those lead to source/sink structures. To take advantage of the potential source/sink structures, we need to know where they are, and the replacement mechanism that allows the source to persist.

The simplest “null model” of larval dispersal assumes that dispersal distance is scaled as the product of the planktonic larval duration and a velocity that represents the diffusive variability in currents. Empirical information on larval dispersal is obtained from four sources: (1) biophysical models of circulation and larval behavior, (2) variation in genetic identity over space, (3) elemental fingerprinting, and (4) direct observations of biological and physical conditions (e.g., time and place of settlement and associated circulation). As outlined below and elsewhere (e.g., Levin 2006), there has been considerable recent progress in these approaches, and results frequently contradict the “null” expectation that dispersal distance is simply related to planktonic larval duration.

Pelagic larval duration (PLD), the amount of time spent in the water column, is perhaps the most obvious biological characteristic influencing the spatial extent of dispersal, with the expectation that dispersal distances increase with PLD (e.g., Siegel et al. 2003). While PLD can be coupled with mean current velocity to obtain dispersal distances (e.g., Roberts 1997), such a simplistic approach is typically not valid and it is generally recognized that metapopulation level outcomes are more complex than this. Dispersal results in complex spatial patterns that reflect an interaction of flow patterns that vary in space and time with patterns in survival and behavior (Largier 2003). True dispersal distances are invariably shorter than simplistic advective estimates as there is a clear tendency to use an average of peak flow speeds rather than a Lagrangian average of water flow velocities taken over appropriate time and space scales (Largier 2003). Gawarkiewicz et al. (2007) discuss some of the physical complexities involved with alongshore transport and cross-shore exchange that will influence larval retention and dispersal, noting that transport will not only change over time, but that travel pathways may occur in multiple phases.

Shanks et al. (2003) compiled information on dispersal distances of marine organisms that had been estimated from (1) direct observation of dispersing propagules, (2) the spatial distribution of larvae, (3) measurements of settlement distances from isolated adult populations, and (4) the spread of introduced species (Fig. 5). The vast majority of the mean realized dispersal distances they report for marine animals are invertebrates; distances for only two fish species are listed (the snapper Lutjanus kasmira 33–130 km and the sculpin Oligocottus maculosus <1 km). The distances vary widely, from <1.0 m for coral planulae to 4,400 km for mollusk veliger larvae. The PLDs for these organisms range from minutes to months. The longest distance dispersers were mollusks and crustaceans, and dispersal distance was positively correlated (r2 = 0.61) with PLD.
https://static-content.springer.com/image/art%3A10.1007%2Fs11160-008-9092-z/MediaObjects/11160_2008_9092_Fig5_HTML.gif
Fig. 5

Estimated dispersal distance plotted as a function of propagule duration. Dashed line is the best fit to the data for the log–log plot. Open circles represent animal populations, closed circles represent plant populations, and circles labeled with letters represent species with dispersal distances that are significantly lower than expected, given their propagule duration. Significant correlations between propagule duration and dispersal distance exist for these data, both with and without the individually labeled data points. From Shanks et al. (2003)

Dispersal models based on 3-dimensional oceanographic circulation models provide dispersal results more consistent with empirical estimates than the PLD null model alone. For example, Cowen et al. (2003, 2006) used ocean circulation and individual-based models that incorporate larval behaviors, PLDs, adult spawning strategies, and larval mortality to estimate that Caribbean-wide reef fish dispersal distances at demographically meaningful time scales are on the order of 50–100 km. Using the same approach for snapper spawning aggregations around Cuba, Paris et al. (2005) suggested that 37–80% of total recruitment is of natal origin. James et al. (2002) and Bode et al. (2006) used circulation modeling to identify key features of the larval connectivity patterns for reef fish larvae in the Great Barrier Reef.

In a more theoretical study related to circulation, Siegel et al. (2003) used stochastic Lagrangian simulations to model dispersal of passive larvae using a range of PLDs in an idealized nearshore region, parameterized with surface velocity statistics typical of coastal currents in central California. The resulting dispersal scales varied from a few km to over 400 km. They also described the interactions between levels of advection and diffusion in providing for self seeding (cf., single reserve persistence above; Largier 2003).

Circulation modeling can result in much more specific results (spatially detailed dispersal kernels) than empirical information sources, but this field is in its nascent phase. The accuracy of projected dispersal patterns will vary, and efforts to cross check and improve accuracy need to be made. Two challenges limit the credibility of present efforts: (1) the ability of these models to correctly represent water motions for both large-scale circulation and small-scale mixing, and (2) the ability to correctly represent typical Lagrangian trajectories, even when larvae exhibit simple behavior. Comparisons with drifter and biological field data should accompany simulations where possible, and methods that assimilate data will likely be more consistently accurate (see Werner et al. 2007 and references therein).

Genetic methods depend on isolation by distance, and they typically try to use “neutral markers” so that post-settlement selection has little influence on patterns of genetic relatedness. Examples indicate a wide range of dispersal distances. Taylor and Hellberg (2003) found a high degree of genetic population differentiation among Caribbean cleaner gobies (Elacatinus evelynae) on the scale of 23 km. On the other hand, Withler et al. (2001) examined microsatellite loci in Pacific Ocean perch (Sebastes alutus) from British Columbia, Canada to determine population structure and revealed three distinct populations: one along the entire west coast of Vancouver Island and two coexisting within Queen Charlotte Sound. Roques et al. (2002) examined microsatellite DNA from deep-water redfish (Sebastes mentella) across the entire North Atlantic to discover a surprising lack of structure across the 6,000 km wide panoceanic zone, but distinct populations in Norway and the Barents Sea, and Gulf of St. Lawrence and offshore Newfoundland. These latter two situations could exist if dispersal and gene flow in these different populations of the same species are of very different scales, most likely determined by differences in their physical settings, e.g., flow regimes, coastal configuration and seafloor topography. On the other hand, this spatial structure could result from selection. This is similar to the case of barnacles along the California Current, which show a latitudinal cline in mitochondrial haplotype, possibly reflecting selection (Sotka et al. 2004). There is increasing evidence of local adaptation, even among pelagic dispersers (Sotka 2005).

In spite of unresolved uncertainties, information from genetic studies has been useful in describing the potential range of dispersal distances. For example, to assess the range in dispersal distance for reserve planning for implementation of California’s Marine Life Protection Act, scientists used results from Kinlan and Gaines (2003) who assembled estimates of dispersal distances using genetic isolation by distance slopes for macroalgae, invertebrates, and fish (Fig. 6). Mean dispersal distances varied widely from a few meters to hundreds of kilometers. Focusing on the 25 species of demersal fish from a variety of environments, estimated dispersal distances range from a few kilometers to hundreds of kilometers, with estimates centered around 100 km. While this figure confirms that we need to plan for a range of dispersal distances we must keep in mind the caveat that “we do not know dispersal distance for most marine species of commercial or conservation interest to within 1–3 orders of magnitude” (Sotka and Palumbi 2006).
https://static-content.springer.com/image/art%3A10.1007%2Fs11160-008-9092-z/MediaObjects/11160_2008_9092_Fig6_HTML.gif
Fig. 6

Mean dispersal distance estimates for benthic marine organisms obtained from genetic isolation-by-distance slopes. From Kinlan and Gaines (2003)

Because exchange of only a relatively few individuals per generation is necessary to render populations genetically indistinguishable, studies of genetic relatedness can tell us if there is little or no connectivity among populations, but reveal little information about higher levels of connectivity (Hedgecock et al. 2007). Genetic studies can provide strong evidence of a lack of movement across specific boundaries. However, they do not provide information on movement at lesser scales within those boundaries, which are scales that can be demographically important. As a result, they will tend to overestimate functional population dynamic dispersal distances because they are sensitive to possible infrequent larval transport over large distances which are functionally insignificant (Hellberg 2006).

Elemental analysis and chemical marking of hard body parts (e.g., otoliths in fish) is currently the most direct way to estimate demographically relevant dispersal patterns. Jones et al. (1999) used tetracycline to make a fluorescent mark on the otoliths of damselfish (Pomacentrus amboinensis) larvae at Lizard Island, Australia and estimated that 15–60% of recruited juveniles were locally produced, having dispersed <10 km. Using otolith microchemistry analysis, Swearer et al. (1999) estimated that ~70% of the recruits of the bluehead wrasse (Thalassoma bifasciatum) to St Croix, U.S. Virgin Islands, were locally recruited, with dispersal distances on the order of 10 s of km. Based upon a combination of otolith microchemistry and genetic (parentage) analysis of a population from a small island north of Papua, New Guinea, about one-third of the settled clownfish (Amphiprion polymnus) juveniles had returned to a 2 ha natal area, many settling <100 m from their birth site (Jones et al. 2005). Miller and Shanks (2004) combined otolith microchemistry and microstructure of black rockfish (Sebastes melanops) along the open west coast of the United States and estimated dispersal distance to be <120 km, roughly two to five times less than the distance based solely on PLD and mean flow (Shanks et al. 2003). These results from otolith microchemistry were corroborated by significant genetic structure detected among adult black rockfish collected 340–460 km apart (Miller et al. 2005). In an invertebrate study, Becker et al. (2007) found two co-occurring congeneric mussels in southern California dispersed on the order of 20–30 km from their origins, but had substantially different patterns of connectivity. The fact that elemental analysis uses natural tags means the number tagged is unknown, which limits the possible conclusions. This effect can be avoided by transgenerational marking of known numbers of adult females which then become incorporated in embryonic otoliths (e.g., Thorrold et al. 2006; Almany et al. 2007). Another potential problem with natural tags is that variability in markers may not exist on the appropriate scales in some locations (Ruttenberg and Warner 2006).

Conclusions from direct observations are difficult because of the challenges involved in tracking larval movement. Nonetheless, we have learned a considerable amount from time-series of postlarval settlement, the sampling of larvae in the plankton, and associated physical circulation conditions. Epifanio and Garvine (2001) review physical and biological mechanisms affecting larval dispersal along the east coast of the United States. Shanks and Eckert (2005) note that mechanisms should differ on the U.S. west coast because of the equatorward and offshore flows in this upwelling system (see Parrish et al. 1981). They propose that the life histories should evolve in a way that mitigates offshore larval loss, and present a review of the larval stages of a number of species in the California Current to demonstrate that. However, there are exceptions, such as larval retention in upwelling shadows. Observations at a number of points along the coast adjacent to the California Current have identified larval retention zones associated with capes and points along the California Current. These retention zones enable population persistence, and also shape dispersal kernels (e.g., Wing et al. 1995a, b, 1998; Graham and Largier 1997; Diehl et al. 2007).

From these different sources of information, we can conclude that there is likely to be a range of dispersal distances, similar to the distribution in Fig. 6. However, we have few precise estimates for dispersal among all locations, and many of these estimates are from topographically complex coral reef environments (e.g., islands) where restricted flow would be expected. This indicates that we should assess reserve designs over a range of mean dispersal distances. It appears unnecessary to assess the effects of shape of symmetrical dispersal patterns (i.e., those without advection; Lockwood et al. 2002). The potential for advection remains a dominant source of uncertainty in projecting sustainability or yield.

Juvenile and adult connectivity

Post-settlement movement of juveniles and adults will influence the amounts of time spent in and out of reserves. The degree to which individuals cross reserve boundaries into fished areas (commonly referred to as spillover [e.g., McClanahan and Mangi 2000; Russ et al. 2003]) affects both yield and sustainability of fisheries.

The many types of movements displayed by juvenile and adult fish pose a significant modeling challenge. The most common method for incorporating juvenile and adult movement into models of MPAs thus far has been diffusion across a boundary between protected and fished areas (an approach originating in Beverton and Holt 1957). Incorporation of more complex movement patterns (e.g., home ranges, ontogenetic shifts, spawning movements) is only beginning to occur in ecological and fisheries models. We first review how juvenile and adult movement can affect sustainability and yield. We then assess what empirical work reveals about juvenile and adult movement.

Effects of juvenile and adult movement on sustainability

Before resorting to formal models, we can conclude that the sustainability or persistence of any specific population with juvenile and adult movement is likely to be less than indicated by models that assume completely sedentary post-settlement juveniles and adults. It is clear that spillover of juveniles and adults increases mortality from fishing. The resulting decrease in FLEP owing to greater mortality will diminish the local tendency for replacement, thus reducing persistence. As a result, MPAs are likely to be most effective in building up a localized spawning stock for species with low rates of juvenile and adult mobility and limited home ranges (Botsford et al. 2003). As the spillover of juveniles and adults across reserve boundaries increases, or reserve size decreases, reserves become less effective, and total population abundance will eventually approach that of a system without no-take reserves, with mortality reduced by the fraction of space in reserves.

The effect of movement rates on the buildup of spawning stock biomass was demonstrated in early, per-recruit models of MPAs. Polacheck (1990) and DeMartini (1993) used YPR and EPR models that assumed cohorts developed in areas that were in part no-take reserves, and that individuals diffused across reserve boundaries at rates determined by the abundance in the reserve and non-reserve areas. In all cases examined, spawning stock biomass per-recruit (SSBR, Table 1) increased with increasing area in reserves, and this effect was greatest when transfer rates over reserve boundaries were low (e.g., a fish with low mobility). The percent increases in SSBR with reserves versus management without reserves were greatest when the fishing mortality rate outside reserves was high. Increased cross-boundary movements impair sustainability because they decrease lifetime spawning opportunities, thereby reducing a population’s reproductive potential (Sladek Nowlis and Roberts 1999; Meester et al. 2001; Gaylord et al. 2005).

Walters et al. (2007) represented diffusion within a specific proposed pattern of MPAs along a coastline using a model with larval connectivity and concentration of effort, but no age structure. Their results indicated that as the strength of a diffusive mechanism increased sustainability could decline substantially. The question then becomes whether juvenile and adults diffuse at the indicated rates, or exhibit more site fidelity by restricting movements within a home range.

Effects of juvenile and adult movement on yield

In models of MPAs that include juvenile and adult movement, there is an inherent tradeoff between sustainability and yield. Even low juvenile and adult movement rates reduce replacement and the buildup of spawning stock biomass inside reserves, with the result of less sustainability and less recruitment. This lower sustainability due to movement outside reserves is accompanied by a possible yield benefit that would not be available if individuals were sedentary, remaining in the reserves. As movement rates increase, yield can rise as fish spend less time in reserves and are more vulnerable to the fishery, but this increased yield comes at the expense of the buildup of biomass, and thus sustainability, in reserves.

There have been a number of theoretical investigations of the effects of juvenile and adult movement on yield from reserves, but because they are cast in terms of effects on YPR (hence do not include effects on recruitment), the effects of movement on yield have not been completely resolved. Because movement affects both EPR and YPR, and change in EPR affects recruitment, the ultimate effect of movement on yield (i.e., recruitment times YPR) is not completely specified by the effects on YPR.

Beverton and Holt (1957) considered the effect of YPR when a certain area of the habitat was not vulnerable to the fishery and movement between the fished and unfished area occurred by random diffusion. In general, closure fractions from zero to 25% produced nearly identical YPR curves. However, large closure fractions (>25%) resulted in some increases in YPR, but only at levels of fishing mortality much higher than those producing MSY in the no-reserve case (see Fig. 1 in Guénette et al. 1998, or Fig. 1a in Gerber et al. 2003). Using elements of Beverton and Holt’s (1957) models, Polacheck (1990) and Demartini (1993) found that YPR was generally lower with reserves, unless the species was fished very heavily, and even then increases in YPR were modest. While these studies of the effects of diffusion between two compartments may provide some hints of population behavior in a system of MPAs, their interpretation is limited because they do not represent the actual spatial configuration of a system of MPAs. In an actual system of MPAs there will be many boundaries between MPAs and fished areas, and when there is some site fidelity, EPR and YPR will vary continuously over space. For example, YPR will be higher and EPR will be lower for a fish with the center of its home range just inside the boundary than for a fish with the center of its home range well inside the boundary.

In addition to the studies assuming diffusive movements between fished and reserve areas, more recent work has incorporated other forms of juvenile and adult movement. Apostolaki et al. (2002) present a detailed model of the yield changes expected for the Mediterranean hake (Merluccius merluccius) fishery where a combination of ontogenetic and spawning movements occurred between nursery and spawning areas. Apostolaki et al. (2002) found that yields are greatly increased for the scenario where the nursery area is a reserve and the primary fishery is for juveniles, over a wide range of movement rates.

Comparison of effects of juvenile and adult movement with available information

Fortunately, the number of species for which we have actual observations of adult/juvenile movement patterns is much greater than the paucity available for larval movement. Movement within a home range is a type of movement that is important to MPAs (Kramer and Chapman 1999). However, a variety of departures from this type of movement behavior have been identified. Examples include spawning migrations (Martell et al. 2000), ontogentic shifts in distribution (Dahlgren and Eggleston 2000), density-dependent movement (Abesamis and Russ 2005), intraspecific variation in movement (Attwood and Bennett 1994), and habitat influences on home range size (Matthews 1990).

Home range is the area in which an individual spends the majority of its time and engages in most of the routine activities of foraging and resting (Quinn and Brodeur 1991; Kramer and Chapman 1999; Pittman and McAlpine 2003). Direct measurement of home range sizes involves marking individuals visually or acoustically and following or relocating them after a biologically relevant time step. Though the extent of individual home ranges differs among and within species, in general there is a positive relationship between body-size and home range size (Fig. 7, Table 2, Kramer and Chapman 1999; Palumbi 2004).
https://static-content.springer.com/image/art%3A10.1007%2Fs11160-008-9092-z/MediaObjects/11160_2008_9092_Fig7_HTML.gif
Fig. 7

Compilation of home range area estimates plotted as a function of mean body length. Many of the coral reef estimates were previously collated in Kramer and Chapman (1999). Other estimates are derived from studies utilizing telemetry data. Species, home range and reference information can be found in Table 2

Table 2

Species, home range areas, and references for data plotted in Fig. 7

Group

Species

Home range (m2)

Reference

Coral reef fish

Amblycirrhitus pinos

3

Kramer and Chapman (1999)a

Bluespine unicornfish

3,717

Meyer and Holland (2005)

Bodianus rufus

449

Kramer and Chapman (1999)a

Canthigaster rostrata

20

Kramer and Chapman (1999)a

Caranx ignobilis

492,250

Wetherbee et al. (2004)

Centropyge argi

1

Kramer and Chapman (1999)a

Cephalopholis cruentata

2,120

Popple and Hunte (2005)

Dascyllus aruanus

3

Kramer and Chapman (1999)a

Enneanectes atrorus

0

Kramer and Chapman (1999)a

Epinephelus guttatus

862

Kramer and Chapman (1999)a

Epinephelus tauvina

344,000

Kaunda-Arara and Rose (2004a, b)

Kyphosus sectatrix

34,423

Eristhee and Oxenford (2001)

Micrognathus ensenadae

21

Kramer and Chapman (1999)a

Ophioblennius atlantius

1

Kramer and Chapman (1999)a

Parupeneus porphyreus

19,201

Meyer et al. (2000)

Plectropomus leopardus

10,458

Kramer and Chapman (1999)a

Pomacentrus flavicauda

2

Kramer and Chapman (1999)a

Sparisoma viride

497

Kramer and Chapman (1999)a

Stegastes diencaeus

3

Kramer and Chapman (1999)a

Stegastes partitus

5

Kramer and Chapman (1999)a

Stegastes planifrons

3

Kramer and Chapman (1999)a

Temperate reef fish

Gadus morhua (juvenile)

60,000

Cote et al. (2004)

Pagrus auratus

55,500

Parsons et al. (2003)

Paralalbrax clathratus

3,349

Lowe et al. (2003)

Red morwong

3,639

Lowry and Suthers (1998)

Sarpa salpa

49,482

Jadot et al. (2006)

Sebastes mystinus

8,783

Jorgensen et al. (2006)

Semicossyphus pulcher

15,134

Topping et al. (2005)

Elasmobranch

Dasyatis lata

1,320,000

Cartamil et al. (2003)

Negaprion brevirostris

680,000

Morrissey and Gruber (1993)

Reptile

Chelonia mydas

16,620,000

Seminoff et al. (2002)

aSee Kramer and Chapman (1999), Appendix 1, for original reference

Marine organisms from a broad range of taxa undergo reproductive migrations from feeding grounds to spawning, mating, or breeding grounds. Spawning movements may be repeated daily for some coral reef fish (Fitch and Shapiro 1990; Mazeroll and Montgomery 1998); annually for many other fish, birds, and mammals, or at up to 9 year intervals for some turtles (Paladino and Morreale 2001). Reef fish tend to be relatively site attached, remaining within a limited home range, but many species make regular migrations to spawning grounds (Claro and Lindeman 2003; Kaunda-Arara and Rose 2004a).

The location or size of an animal’s home range may change during its lifetime as the result of ontogenetic development (Pittman and McAlpine 2003; de la Morinière et al. 2002). Home ranges may also temporarily shift in response to short term environmental changes such as storms or seasonal changes. MPAs will have to be designed to account for these changes, and greater area will be required than for any single stage.

Density dependent movements are particularly relevant for MPAs since reserves frequently increase population abundance and biomass (Halpern 2003). In territorial species, the losers of disputes, which tend to be smaller, younger, and or lower quality individuals, may relocate to less suitable habitat or remain in an area as “floaters” (Quinn and Brodeur 1991; Kramer and Chapman 1999). Conversely, in non-territorial species it is more likely that the older and/or healthier individuals are the ones to disperse, since they have the speed, stamina, and energy reserves necessary for a successful relocation (Quinn and Brodeur 1991). Density dependent changes in fish movement have not been directly evaluated in modeling studies, but it is likely that they will work in opposition to the desired increases in density intended with the implementation of MPAs.

In addition to variability among species in the scale and patterns of movement, there is often variability within species as different individuals exhibit polymorphic movement behavior (Quinn and Brodeur 1991; Attwood and Bennett 1994; Willis et al. 2001; Starr et al. 2004). Some individuals of essentially site-attached species disperse relatively long distances to new home ranges (Quinn and Brodeur 1991; Kaunda-Arara and Rose 2004a). An extreme example was described for Galjoen (Coracinus capensis), a surf zone dwelling fish (Attwood and Bennett 1994). The majority of Galjoen were estimated to have home ranges on the order of a few kilometers, while some individuals were recaptured more than 1,000 km from the original tagging location.

Habitat quality can influence the movement and home range sizes and locations of marine organisms. Individuals occupying lower quality habitat, with lower densities of food and prey, are likely to range more widely than those whose home ranges include high-quality habitat (Quinn and Brodeur 1991; Kramer and Chapman 1999; Christensen et al. 2003; Laurel et al. 2004). Matthews (1990) found that both the size of home ranges, and the probability of making seasonal movements, differed as a function of habitat relief for three species of rockfishes. In contrast, movement may decrease in less suitable habitat. For instance, in field and laboratory studies, two species of cod Gadus morhua and G. ogac aggregated more tightly over lower quality sandy substrate than over eelgrass (Laurel et al.2004).

Habitat structure can also provide corridors or pathways for movement. Movement of bluespine unicornfish (Naso unicornis) within home ranges on a Hawai’ian reef corresponds with topographic habitat features (Meyer and Holland 2005). Spawning migrations of the female blue crab (Callinectes sapidus) occur along a deeper-water channel within the shallower waters of Chesapeake Bay (Lipcius et al. 2001). Long distance movement by three species of coral reef fish on a Kenyan reef were likely facilitated by the shape of the reef, which had a long along-shore dimension (Kaunda-Arara and Rose 2004a). In addition, habitat structure can affect movement by acting as physical or behavioral barriers (Chapman and Kramer 2000).

Modeling of the effects of juvenile and adult movement has focused mainly on the effects on YPR and EPR of diffusion between reserve and non-reserve areas, and has not yet begun to evaluate the effects of detailed variation in the dominant type of behavior, home ranges (but see Kramer and Chapman 1999; Meester et al. 2001). It is clear from studies thus far that variation in juvenile and adult movement among resident species will have varying effects on the efficacy of MPAs. Because of the greater ease of tagging and tracking juveniles and adults, we have much more information on movement in this phase of life histories. However, modeling indicates that pure diffusion, relative to more restricted home range movements, can have a substantial negative effect on sustainability and yield (Walters et al. 2007). Hence, more effort is needed in analyzing and categorizing the various forms of movement to see how close they come to the random independent movements that underlie the assumption of diffusion and lead to constant flux across differences in abundance.

Economic connectivity

The effect of the third type of connectivity, the movement of fishermen, on sustainability and yield is substantial, hence these effects need to be added to those of larval dispersal and juvenile/adult movement to achieve a comprehensive understanding of connectivity in MPAs. Here we are concerned with entry and exit behavior of fishermen within a fishery, as well as changes in spatial fishing patterns when MPAs are used in management.

The decisions regarding when to participate in a fishery and the choice of where to fish are governed by similar economic forces (Wilen 2004). In general, the perceived relative profitability of fishing determines both kinds of decisions. In an open access system, new fishermen will enter the fishery so long as the profit that can be gained by fishing exceeds profits that can be attained elsewhere (Gordon 1954). The effect of increased entrants into the fishery is decreasing profitability in the resource and the equilibrium state will be a level of profit equal to the profit level available from outside (other than fishing) opportunities. This theory can be extended to the choice of fishing location in the sense that the choice of fishing areas will be a discrete choice between multiple areas with different profitability. As such, fishermen will be likely to fish in the area with the highest relative profitability, with movement between the most profitable patches occurring over time (Sanchirico and Wilen 1999).

In the fisheries and economic literature, a variety of methods are used to describe the sustainability and yield consequences of fishermen’s movements in the context of MPAs. Some representations do not explicitly include economics but rather make simple assumptions about changes in the spatial distribution of effort caused by MPAs (e.g., Beverton and Holt 1957; Holland and Brazee 1996; Halpern et al. 2004; Kaplan and Botsford 2005). Other papers explicitly include economics and tend to be more focused on the fishery sector rather than the managed population (e.g., Holland and Sutinen 1999; Smith and Wilen 2003, 2004; Hicks and Schnier 2006). Both classes of studies provide insights as to how MPAs can perform as a fishery management tool.

Non-economic assumptions of shifts in fishing effort with reserves

MPAs can be viewed as a form of effort control, if indeed the fishing effort that existed within reserve areas is eliminated when reserves are instituted. On the other hand, reserves can also serve to concentrate fishing effort into a smaller habitat area if fishermen do not leave the fishery, but rather move to non-reserve areas. Beverton and Holt (1957) provide the earliest example of how YPR is affected by reserves where the total fishing effort is fit into the available, smaller fished area. For species that are heavily exploited, MPAs can provide increases in YPR, though appreciable increases are only seen when a large fraction of the total area is designated a reserve (e.g., >25%). Many others (Holland and Brazee 1996; Guénette et al. 1998; Apostolaki et al. 2002; Botsford et al. 2003) have noted the generality of this result and have posited that a reserve can have a similar effect to increasing the age of first capture.

Such a per-recruit analysis ignores the effect of MPAs on recruitment and the consequent increases or decreases in yield that reserves are likely to produce. The essential question then becomes whether the increased production from reserves donated to fished areas can compensate for the high levels of fishing effort outside reserves resulting from effort redistribution. Halpern et al. (2004) proposed that increased egg production from reserves could outweigh the effects of redistribution of effort resulting in a net fisheries yield benefit. Their model assumed that fishing effort was uniformly distributed in the fished areas, larval dispersal was the only form of movement, the dispersal was of the “larval pool” form, and larval mortality was density-independent, i.e., larval settlement was linear function of larval production. If biomass was three times greater in reserves relative to fished areas [(a quantity based on MPA meta-analysis results in Halpern (2003)], the model predicted that the increased egg production of the population within reserves would compensate for the increased fishing effort concentrated into the fished area.

In a more detailed simulation model, Kaplan and Botsford (2005) explicitly compare yield for three fishery cases: (1) no marine reserve where fishing mortality was evenly distributed over space, (2) 20% of habitat in MPAs and effort in reserves disappears with reserve formation, and (3) 20% of habitat in MPAs and effort is redistributed into fished area after with reserve formation. Very generally, adding fishermen movement by a simple redistribution of effort essentially just increases the fishing mortality rate, resulting in a smaller advantage of MPAs over conventional management, regardless of larval dispersal distance.

Each of the studies discussed to this point assume that fishing effort is uniformly distributed in the areas open to fishing. However, as mentioned previously, fishermen do not uniformly or randomly distribute themselves over fishing grounds. Hilborn et al. (2006) investigated this effect by incorporating an aggregation parameter that controlled the degree to which simulated fishermen would be attracted to high abundances. Because they used logistic population models in which individuals at all stages (larvae, juveniles and adults) are identical, and movement was a normal redistribution kernel meant to represent both larval dispersal and juvenile and adult swimming movement, it is difficult to relate their results regarding fishermen movement to other forms of movement. However, it should be noted that for fish with low mobility, their level of aggregation did make a large difference in yield with MPAs. Kellner et al. (2007) demonstrated that “fishing the line,” the aggregation of fishermen at reserve boundaries, is an optimal fishing strategy when MPAs are a part of management. Fishing the line strategies can increase catch over a uniform distribution of fishing effort outside MPA boundaries under many possible scenarios.

Economic decisions and spatial distributions of fishing effort

In their modeling investigations of the effects of fishermen’s movement behavior, economists tend to focus on the fishermen, and represent the biological populations in a way that does not differentiate between individuals at different life stages at each location, hence they cannot be tied to critical observable biological rates. The logistic models used have only a single state variable at each location, which presumes that all individuals (or units of biomass) are identical in their contribution to growth, mortality and reproduction. These models have no explicit observable population rates such as fecundity and mortality versus age, hence they could not be used to evaluate the interactions between different values of FLEP and aspects of fisher behavior. Such models demonstrate some of the possibilities regarding fishermen’s behavior, but it is difficult to relate them to the two other forms of connectivity (e.g., to ask how changes in recruitment relationships and fisher behavior interact to determine sustainability). As examples, Sanchirico and Wilen (1999) and Sanchirico et al. (2006) used logistic models to suggest interactions between various kinds of connectivity.

More specific models of fleet dynamics have been constructed, estimated and analyzed with a goal of understanding how the spatial behavior of fishermen will influence marine reserve function. Holland (2000) developed an integrated biological and economic model that analyzed the effect of various closures on Georges Bank on the SSB, harvest, and revenues of the multispecies New England groundfish fishery. The biological models for the several species modeled were age structured, and spatially distributed over Georges Bank, with the density-dependence in recruitment calculated for the species metapopulation as a whole, and larval pool dispersal. Adult movement was represented by a diffusion term and a seasonal term based on known seasonal movement. The model used to represent movement of fishermen was estimated from empirical data assuming a hierarchy of decisions: a zone choice followed by choice of area within that zone on a trip-by-trip basis. Many of the results could not have been arrived at without incorporating fisherman movement into a biological model. For example, Holland (2000) found that closures on Georges Bank would lead to increased equilibrium SSB for many species on Georges Bank, but could result in decreased SSB for stocks in adjacent areas (Gulf of Maine and Southern New England) due to increased fishing effort in those areas. Harvests with reserves implemented in Georges Bank could increase or decrease, depending on the species. But again, displaced effort to adjacent areas off Georges Bank could lead to long term decreases in harvest in these areas due to overfishing. The results were quite sensitive to assumptions about biological connectivity, represented by both seasonal migrations and diffusion between adjacent fishing zones.

Smith and Wilen (2003, 2004) fit a similar model of fishermen’s choice to data from the northern California red sea urchin fishery, and then coupled that to the same biological model used in Fig. 3, but with different parameter values. They compared their model to the same biological model with uniform effort over space. The equilibrium state of both models shared the same value of catch, but the model with fishermen behavior (i.e., without uniformly distributed fishing effort) had roughly three times the steady-state egg production. The unequal distribution of effort essentially caused the system to behave like an aggregate population with lower fishing mortality, which increased EPR without decreasing YPR, and may have increased total recruitment. They conclude there is little or no system wide economic benefit to MPAs since long term harvest benefits of the reserve were outweighed by the short term losses incurred immediately following reserve creation. In a similar study, Hicks and Schnier (2006) found that the costs of MPAs may be lower if one considers a model where fishermen select an optimum economic trajectory of fishing locations that maximize their discounted revenues over the course of a fishing trip.

In summary, the simple approach of effort redistribution after MPA institution leads to the interpretation that the effects of larval dispersal and juvenile and adult movement on sustainability and yield will be the same as identified in previous sections, but will correspond to higher fishing effort, hence there will less difference between conventional and spatial management. The models that explicitly depict fishermen’s choices show a much more detailed and subtle redistribution of effort that can be affected by many factors, such as distance from port, market price, and weather. In many of these studies, MPAs achieve sustainability goals by effectively raising costs, and lowering the efficiency of fishing (Hannesson 1998).

Synthetic conclusions

The studies reviewed here reveal an evolving understanding of the effects of larval, juvenile/adult and economic connectivity on sustainability and yield in spatial management of fisheries. Results from modeling studies indicate the important aspects of each of the three kinds of connectivity. Comparison of those with results from empirical studies of the biophysical mechanisms underlying each type of connectivity leads to an appreciation of the uncertainties currently involved in the spatial management of fisheries. These can be compared to the various uncertainties involved in conventional fishery management.

In response to concerns regarding overfishing, conventional management has increased its focus on sustainability as a goal worth pursuing directly and tracking. The various reference points associated with sustainability reflect population abundance and population growth rate, and the latter is frequently associated with the familiar population dynamic concept of replacement through FLEP.

In conventional fishery management, uncertainty in the state of the population is accounted for, when possible, by control rules which indicate action to be taken when combinations of estimated population abundance and/or population replacement rate fall below pre-determined thresholds. As noted above, the replacement level is often represented by the value of fishing mortality rate that corresponds to a specific value of SPR or FLEP. The abundance can be chosen somewhat arbitrarily, based on acceptable risk, but the replacement rate should be greater than the CRT. The CRT can sometimes be estimated by estimating the steepness of the stock-recruitment curve, but for most fisheries it is not known. The value of the CRT, which fundamentally determines how much fishing can be tolerated without collapse, is an important uncertainty in fishery management.

Explicitly accounting for the spatial nature of fisheries requires that replacement be computed through all possible pathways over space, which depends on: (1) the spatial distribution of FLEP over space and (2) the larval dispersal pattern from each location. We can use an example of the spatial distribution of FLEP to illustrate how the three different types of connectivity are related to sustainability. For species with sedentary adults, the FLEP distribution is determined by the local intensity of fishing and the distributions of the benthic habitat and MPAs (Fig. 8a, also as in Fig. 2). Adding an account of the effects of movement of the fished juvenile or adult stages using home range behavior smoothes the distribution of FLEP near the edges of the MPAs owing to fish with home ranges near the boundaries crossing them (Fig. 8b). This has relatively little effect on sustainability in this example, primarily because the MPAs are for the most part larger than the home range size, 6 km. If the juvenile/adult movement were not site based, but rather of a purely diffusive nature, the smoothing effects of the FLEP distribution would be greater. Finally, explicitly adding the effects of economic connectivity to the problem will lead to heterogeneity in the distribution of effort, hence fishing mortality, due to spatial variability in costs of fishing (e.g., the increased cost of traveling farther from port). This changes the distribution of FLEP as shown in Fig. 8c, where a port is assumed to exist at 325 km.
https://static-content.springer.com/image/art%3A10.1007%2Fs11160-008-9092-z/MediaObjects/11160_2008_9092_Fig8_HTML.gif
Fig. 8

A comparison of the effects of the three types of connectivity on hypothetical spatial distributions of FLEP, and their consequences for sustainable settlement rate. (a) The effects of larval dispersal only, (b) the effects of larval dispersal and adult/juvenile movement within a home range of 6 km, (c) the effect of larval dispersal, juvenile/adult movement, and a hypothetical dependence of fisher movement on travel costs from a port at 325 km. Settlement was computed using the Dispersal per Recruit (DPR) method (Kaplan et al. 2006) with a hockey-stick settlement-recruit function and critical replacement threshold of 35%

Translating each FLEP distribution in Fig. 8 to a spatial distribution of sustainability (e.g., equilibrium settlement rates) requires knowing or assuming larval dispersal patterns. Considering larval connectivity only (Fig. 8a), computation of replacement for symmetrical dispersal patterns, and density-dependent recruitment that is monotonically increasing with settlement, indicates that short distance dispersers will be sustained in single reserves with width equal to at least their dispersal distance, and that longer distance dispersers will be sustained on sections of the coast where a specific fraction of the coast is in reserves. For high fishing rates, that fraction is the CRT (e.g., 35–50%) for the species, and the fraction declines as the intensity of fishing declines. The presence of advection reduces the likelihood of sustainability. Under these recruitment assumptions, yield with MPAs will likely not be greater than yield through conventional effort controls, unless F is greater than FMSY, and yield from short distance dispersers in isolated small reserves will be limited. Yield could be about the same as that possible with conventional management, and the maximum possible yield will come from a large number of small reserves covering the minimum fraction of coastline required for sustainability (Hastings and Botsford 1999, 2003). Yield can be greater than that possible through conventional management when there is a decoupling of local density-dependence and reproduction, such as under strong source-sink conditions (Morgan and Botsford 2001; Gaines et al. 2003), and when settlement is limited by local adult density (Gaylord et al. 2005; Ralston and O’Farrell 2008; White and Kendall 2007).

A qualitative assessment of some uncertainties involved in management with MPAs follows from this description. Many aspects of sustainability and yield depend on dispersal distance, yet our assessment of knowledge of dispersal patterns above concluded that, while there is likely to be a range of distances, we have extremely limited knowledge of dispersal patterns. For a species whose larvae disperse short distances, sustainability is likely and yield increases will be limited to the area near the edges of reserves. For species with longer dispersal distances, sustainability will depend on the level of fishing (i.e., the FLEP) outside the MPAs, and on the value of the CRT for that species, the same uncertain parameter that confounds conventional fishery management. To these sources of uncertainty we need to add the uncertain levels of advection, source-sink structure and density-dependent recruitment.

The effect on replacement calculations of adding juvenile and adult movement to the FLEP distribution (Fig. 8b) depends on relative spatial scales. Because variability in the distribution of MPAs is being smoothed by a weighting function that depends on home range size, the overall impact of juvenile/adult movement depends on the scale of spatial variability in FLEP relative to the scale of juvenile/adult home range. If the home range is small relative to the mean width of MPAs and the spatial scale of variability in habitat, the effect on the shape of the realized FLEP will be relatively small. The effect on sustainability will always be to reduce it, and the magnitude of that effect will increase with the scale of the home range relative to the average size of the MPAs. The relative impact of juvenile/adult movement on yield will increase with the spatial scale of that movement, relative to the spatial scale of larval dispersal of persistent species. The effects of pure diffusion could be substantial, and can be surmised by viewing it as extremely large uniform home range. Our assessment of the knowledge of juvenile and adult movement indicated it is better known than larval dispersal, and that for many species it involved site fidelity. To the extent that these conclusions hold, the relative effects of juvenile/adult connectivity on uncertainty in spatial management in fisheries could be relatively low.

The effect on replacement calculations of adding explicit movement by fishermen in response to their local economic conditions (e.g., travel costs) is to redistribute effort, possibly increasing or decreasing total effort (Fig. 8c). Accounting for spatial differences in effort could identify locations where populations are less heavily exploited as well as locations where populations are more heavily exploited. The increased variability in costs and benefits of MPAs in various areas, and their interdependence, can increase the complexity of the decision making process considerably. For example, it may turn out that there is such great spatial variability in effort that there are essentially de facto MPAs, and they merely need to be formalized (Smith and Wilen 2003). In our review above, we concluded that while there are relatively few examples of completely quantified movement of fishers, there were not the technical problems in observing that movement that there were in larval and juvenile/adult connectivity. Thus, this is a source of uncertainty that can be reduced by giving it sufficient attention.

The structure of ecosystems within reserves will be determined by the effects of connectivity on replacement. For example, when higher trophic levels are effectively released from fishing mortality, there is a potential for trophic cascades, such as that observed for fishes, urchins, and algae (Sala et al. 1998; Shears and Babcock 2002). Alternatively, diverse prey guilds may weaken cascading effects, as for the case of Nassau grouper predation and parrotfish grazing reported in Mumby et al. (2006). To date, we have relatively little ability to predict cumulative community effects of reserves. Moreover, many ecological studies within reserves implicitly assume that the abundances of resident species, relieved from any fishing mortality, are due entirely to local community interactions. Unfortunately, this assumption ignores the potential effect that reserve design itself, by differentially mediating the connectivity processes of individual species, may have on community dynamics. Ultimately, since the effects of reserves on sustainability will vary with dispersal distance and reserve configuration, it is doubtful that relative species abundance in reserve communities will be the same as with pristine communities.

Both population connectivity within and among reserves can influence the nature of community interactions through their effects on replacement in various species. As discussed above, single reserves will differentially protect species from fishing mortality depending on the scale of dispersal, with species having dispersal distances smaller than the size of the reserve being the biggest beneficiaries of protection. In contrast, more mobile species, to the extent that they “leak” outside of reserves and are caught in fisheries, will experience less benefit from reserves, weakening replacement. Community structure in these reserves should therefore vary as a result of both interaction- and dispersability-based shifts in abundance of community members. In other words, differential connectivity of various components of such communities should influence the outcomes of community interactions (Walters 2000).

Future challenges

Perhaps the most serious challenge for the design of MPA networks that have the best chance of favorably contributing to sustainability and yield of fisheries is the paucity of empirical information on how they work at the population level. There have been a number of reviews of the fraction of some 70 known MPAs in which various attributes such as biomass, mean age and diversity increase (e.g., Halpern 2003), but little focus on why they increase in some, but not in others (e.g., Micheli et al. 2004). Furthermore, effects on fishery yield have been reviewed for only small fraction of these MPAs (e.g., nine in Worm et al. 2006). However, the effects of choices that have to be made in the design of MPA networks (e.g., location, size and spacing) on their eventual success in improving sustainability or yield have received little empirical attention.

There is also a need to reduce uncertainty at the level of the mechanisms of connectivity. Knowledge of larval dispersal scales is especially important for long distance dispersers, and knowing advective effects is critical. Juvenile and adult movement is better known because of a greater ability to observe these movements, but the degree to which there is site fidelity, as opposed to pure diffusion, is critical. There are not the same technical impediments to more information on fisher movement, but there are relatively few quantitative examples, and more are needed. We note that conventional and spatial management share the fundamental uncertainty in CRT, hence that is not a source of uncertainty that is avoided in going to management by MPAs (cf., Lauck et al. 1998).

At this juncture, when MPAs are being recommended globally, it is time for a shift in focus in both empirical observations and modeling from the more strategic questions of what can MPAs do, to how do they do it. Because empirical observations have been cast in terms of questions such as what is the average increase in biomass in MPAs, we have a poor understanding of the effects of connectivity on MPAs and have had to answer related questions using models. It is time to shift focus of empirical studies to questions such as, in what fraction of MPAs did biomass increase, and why did it increase or not in each treatment? This will reduce the number of model assumptions on which MPA design depends, and lead to more empirically based design. The conservation value of an MPA can be dramatically increased by implementation of systematic monitoring and the consequences for future design.

There remain modeling challenges in each of the three types of connectivity. Further understanding of the dynamic implications of temporal and spatial variability in dispersal will improve our understanding of larval connectivity. In juvenile and adult connectivity, home range behavior is apparently the most relevant to outcomes, but it has not been thoroughly investigated through modeling. In the area of economic connectivity, more examples of applications of existing techniques would improve our general understanding of its effects. The greatest modeling challenge is to understand the interactions between these three types of connectivity, and we hope that this review has contributed to this.

The models required for comprehensive study of the interactions among the three types of connectivity will require a consistent level of realism (sensu Levins 1966). They need to explicitly represent the actual mechanisms of movement in the various life stages and in economic factors affecting the distribution of fishing mortality. We need to be able to compare mechanisms in each different type of connectivity in a clear fashion, unoccluded by model differences. For this reason, models that assume all individuals or all units of biomass contribute to dynamics equally (i.e., logistic models) will not be as useful as age-structured models with explicit stock-recruitment mechanisms. This particular requirement will strain our future ability to construct ecosystem models that link the single species described here, perhaps in a way that describes the existing uncertainties.

Finally, as noted elsewhere, some of the tools currently used in the more conventional approach to management will not perform well with spatial management (Field et al. 2006). We will need to employ the emerging understanding of spatial population dynamics to forge a new approach. Perhaps reference points such as SPR (FLEP) can be replaced by a weighted sum or spatial convolution over the spatial distribution of FLEP where the weighting represents connectivity. The question of whether the biomass in reserves should be “on the table” when assessing the status of a population was raised by Field et al. (2006). We would say that the answer is yes, and that the way to assess it would be some version of the sustainability assessment described here (e.g., Fig. 8). This explicit consideration of space will involve an increase in complexity and difficulty, but will account for actual functional structure and uncertainty more directly.

Acknowledgements

This manuscript is a product of an effort to integrate the science of marine protected areas and fisheries management, sponsored by NOAA Fisheries, Southwest Fisheries Science Center and the National Marine Protected Areas Center. We thank Lisa Wooninck and other working group participants for their guidance and feedback.

Copyright information

© Springer Science+Business Media B.V. 2008