Larval Transport and Dispersal in the Coastal Ocean and Consequences for Population Connectivity

M A N Y M A R I N E S P E C I E S have small, pelagic early life stages. For those species, knowledge of population connectivity requires understanding the origin and trajectories of dispersing eggs and larvae among subpopulations. Researchers have used various terms to describe the movement of eggs and larvae in the marine environment, including larval dispersal, dispersion, drift, export, retention, and larval transport. Though these terms are intuitive and relevant for understanding the spatial dynamics of populations, some may be nonoperational (i.e., not measurable), and the variety of descriptors and approaches used makes studies difficult to compare. Furthermore, the assumptions that underlie some of these concepts are rarely identified and tested. Here, we describe two phenomenologically relevant concepts, larval transport and larval dispersal. These concepts have corresponding operational definitions, are relevant to understanding population connectivity, and have a long history in the literature, although they are sometimes confused and used interchangeably. After defining and discussing larval transport and dispersal, we consider the relative importance of planktonic processes to the overall understanding and measurement of population connectivity. The ideas considered in this contribution are applicable to most benthic and pelagic species that undergo transformations among life stages. In this review, however, we focus on coastal and nearshore benthic invertebrates and fishes. Larval Transport and Dispersal in the Coastal Ocean and Consequences for Population Connectivity

M A N Y M A R I N E S P E C I E S have small, pelagic early life stages. For those species, knowledge of population connectivity requires understanding the origin and trajectories of dispersing eggs and larvae among subpopulations. Researchers have used various terms to describe the movement of eggs and larvae in the marine environment, including larval dispersal, dispersion, drift, export, retention, and larval transport. Though these terms are intuitive and relevant for understanding the spatial dynamics of populations, some may be nonoperational (i.e., not measurable), and the variety of descriptors and approaches used makes studies difficult to compare. Furthermore, the assumptions that underlie some of these concepts are rarely identified and tested. Here, we describe two phenomenologically relevant concepts, larval transport and larval dispersal.
These concepts have corresponding operational definitions, are relevant to understanding population connectivity, and have a long history in the literature, although they are sometimes confused and used interchangeably. After defining and discussing larval transport and dispersal, we consider the relative importance of planktonic processes to the overall understanding and measurement of population connectivity. The ideas considered in this contribution are applicable to most benthic and pelagic species that undergo transformations among life stages. In this review, however, we focus on coastal and nearshore benthic invertebrates and fishes.

Larval Transport and Dispersal in the Coastal Ocean and Consequences for Population Connectivity B Y J E S Ú S P I N E D A , J O N AT H A N A . H A R E , A N D S U S P O N A U G L E
Larval transport is defined as the horizontal translocation of a larva between points x 1 ,y 1 and x 2 ,y 2 , where x and y are horizontal axes, say, perpendicular and parallel to the coastline. In larval transport, only the spatial dimensions matter. Although this definition ignores the vertical axis (z) for simplicity, this dimension is critical for larval transport because larvae can modify their horizontal distribution by swimming vertically, thereby encountering different currents (Nelson, 1912;Crisp, 1976). To transfer from point x 1 ,y 1 to point x 2 ,y 2 , a larva can swim horizontally and may be transported by diffusive and advective processes (Scheltema, 1986). Defined as the translocation of a larva between two points, larval transport appears deceptively simple. However, the wide range of larval behaviors and physical mechanisms, together with their variability at multiple scales, makes larval transport exceedingly difficult to measure. The temporal and spatial scales of variability are enormous (Scheltema, 1986), even when considering a single physical transport mechanism (see Box 1).
In contrast, larval dispersal refers to the spread of larvae from a spawning source to a settlement site. This definition is consistent with the terrestrial literature (natal dispersal in Clobert et al., 2001;Begon et al., 2006) that describes seed dispersal as the probability density function of the number of seeds versus distance from the adult source (i.e., the dispersal kernel) (Nathan and Muller-Landau, 2000; see Gerrodette, 1981, for a rare marine example). Using the dispersal kernel, dispersal can be viewed as a probability that a released zygote will make it to settlement over is thought to be a key process for population replenishment, genetics, spread of invasive species, and other phenomena (Cowen et al., 2006, this issue;Levin, 2006

L ARVAL TR ANSPORT Reconsideration of the Scales of Larval Transport
The term larval transport brings to mind small, passive larvae being moved throughout the ocean by meso-and large-scale physical processes (Johnson, 1939). This view has become a paradigm-larvae are released, transported by mesoscale processes, mixed in a larval pool, and then randomly recruited to juvenile or adult habitat (e.g., Roughgarden et al., 1988;Siegel et al., 2003). An increasing number of studies, however, conclude that a significant amount of self-recruitment occurs in marine populations Almany et al., 2007). These conclusions are not in and of themselves surprising: a population is defined as a self-sustaining component of a species, and thus self-recruitment is a defining attribute of a population (Sinclair, 1988). What is surprising is the relatively small spatial scales over which self-recruitment has been observed. For example, despite a planktonic stage of 9-12 days, approximately 30% of settling panda clownfish self-recruited to an area of 0.5 km 2 . The implication of this and similar observations, combined with recent modeling and genetic studies (Cowen et al., 2000;Gerlach et al., 2007) e movement of larvae in internal bores is an example of the variety of spatial and temporal scales involved in larval transport. Larval accumulation at surface-propagating convergences is critical for effective transport in internal bore warm fronts, and the time scales of these convergences are from a few seconds to a few hours. On the other hand, water-column stratification, a seasonal phenomenon, modulates the energy of internal bores and therefore also impacts larval transport (Pineda and López, 2002). At even larger scales, stratification and internal bores are modulated by El Niño, an interannual phenomenon (Zimmerman and Robertson, 1985). us, temporal scales relevant for understanding larval transport by internal tidal bores range from seconds to years. Other temporal scales important to internal tidal bore larval transport that are not depicted here include fortnightly periodicity (~ 14.4 days), and the periodicity of coastally trapped waves (a few weeks; Pineda and López, 2002). In the literature, larval transport generally encompasses horizontal distances ranging from tens to hundreds of kilometers, a usage we follow in this contribution.
1 In this contribution we use the term nearshore to describe (a) the shallow waters where surface and bottom Ekman layers interact, the nearshore of Mitchum and Clarke (1986), and the inner shelf of Lentz (1995), and (b) the surfzone, while the coastal region includes mid-and outer-shelf areas.

Seasonal stratification (months)
Accumulation in internal tidal bore warm fronts (seconds to hours) and the constrained nearshore larval distributions of littoral species (Barnett and Jahn, 1987;Tapia and Pineda, 2007), is that the spatial scales of larval transport may be much smaller than previously recognized. These results indicate that small-scale and nearshore physical processes play an important role in larval transport (Kingsford, 1990;Willis and Oliver, 1990;Pineda, 1999 1989). Salinity (Thièbaut et al., 1992) and water-column stratification (Pineda and López, 2002) contribute to larval transport because sharper stratification in shallow waters (e.g., Hickey, 1979) allows larvae of coastal species to exploit vertically sheared flow to control horizontal distributions (Paris and Cowen, 2004), and internal motions such as internal tidal bores may transport larvae onshore. Surface waves that break near the shore produce some mass transport, and storm systems that originate in the deep ocean sometimes move onshore.
Flows in the nearshore are broken by coastline topographic features such as bays and capes, resulting in complex flows with smaller spatial coherence (see discussion in Okubo, 1994). This is true for cross-shore coastal flows, whose coherence scales are much smaller than the alongshore coastal flows (Brink, 1999). The relative importance of these processes varies with depth and distance from the shoreline (e.g., Lentz et al., 1999;Largier, 2003).

Modulation of Nearshore Cross-Shore Transport by Large-Scale Processes
Clearly meso-and large-scale processes affect larval transport, and most studies emphasize these effects. Large-scale physical processes also influence the smaller-scale processes discussed above.  (Simpson, 1984;Zimmerman and Robertson, 1985

Small-Scale Processes and Event-Type Larval Transport
Spatial and temporal scales are linked in the ocean (Stommel, 1963), so the importance of small-spatial-scale processes underscores the significance of . Relationship between the spatial and temporal components of larval transport, larval dispersal, and reproductive population connectivity for a sessile species. Survivorship is not depicted. Note that the sum of larval transport distances can be larger than the dispersal distance. White circles are locations in space with coordinates x-y at times t. All locations are pelagic except x o ,y 0 and x 4 and y 4 , which are benthic. Distance could also be represented in two dimensions (e.g., x,y as cross-and alongshore axes.) small-temporal-scale processes to larval transport. Moreover, meso-and largescale processes can exhibit small-temporal-scale variability (Stommel, 1963) and be episodic (e.g., hurricanes). Larval settlement from the plankton for many marine organisms is episodic, and it is not uncommon to have the majority of a season's settlement occur in a handful of days Sponaugle et al., 2005).

Behavior and Larval Transport
As our appreciation of small-scale physical processes grows, so does our appreciation for the role of larval behavior in influencing larval transport. For many years, larvae were considered planktonic, that is, moving at the whim of ocean currents but using feeding and predator avoidance behaviors that resulted in small-scale (millimeters to centimeters) movements (Blaxter, 1969). The view of passive larvae gave way to the con-cept that vertical swimming behavior, changes in buoyancy, and ontogenetic changes in vertical position influence the horizontal movement of larvae; this view was adopted early in estuarine and coastal lagoon systems (Nelson, 1912;Pritchard, 1953;Bousfield, 1955) and later in shelf and open-ocean systems (Kelly et al., 1982;Cowen et al., 1993).
Additionally, the influence of larval settlement behavior on the specific location of settlement, at scales of meters to tens Connectivity = ƒ(larval dispersal, post-larval survival)

Larval behavior
Advection, diffusion Dispersal = ƒ(larval transport, survival, spawning and settlement) Larval transport = ƒ(physical transport, larval behavior) Figure 2. e concepts of larval transport, larval dispersal, and reproductive population connectivity. Colors of arrows distinguish each concept. For example, the green arrow in the connectivity box means dispersal is involved in reproductive population connectivity.

28
Larval transport in nearshore and shelf species is often split into crossand alongshore components (e.g., Hare et al., 1999;Ma and Grassle, 2005). is distinction follows a convention in coastal physical oceanography and is convenient because cross-and alongshore hydrodynamic processes have different temporal and spatial scales (Winant, 1983), different physical processes dominate cross-and alongshore transport (e.g., Winant and Bratkovich, 1981), and momentum balances in these two axes are accounted for by different terms (e.g., Lentz et al., 1999).
Also, plankton patches have widely different dimensions in the two axes (Mullin, 1993). Because the strongest gradients in water properties and ecological variables are in the cross-shore dimension, transport on this axis has a disproportionately large effect on the distribution of larvae.
For nearshore species whose later developmental stages move progressively offshore with time, such as the southern California barnacle nauplii (Tapia and Pineda, 2007), cross-shore transport is the most critical process, as older larvae tend to be farther away from the shore and must return nearshore to settle and reproduce. Similarly, for species that move offshore to spawn but have nearshore settlement habitats, such as Atlantic menhaden , larvae must move onshore to recruit to juvenile habitats. Although cross-shelf transport is often emphasized in studies of larval transport, it is obvious that alongshore processes also play a role (Hare et al., 1999), particularly in population connectivity. Nearshore and coastal marine populations are generally arrayed along coasts, and the alongshore movement of larvae between these populations can keep these geographically isolated populations connected.
More recent research shows that larvae also have horizontal swimming capabilities that improve with development (see review by Leis, 2006). For example, larvae of a damselfish swam continuously for 39 hours without food, covering a distance equivalent to 19 km (Stobutzki, 1997). Similarly, larval lobsters and early pelagic stages of cephalopods are good swimmers (Villanueva et al., 1996;Jeffs and Holland, 2000).
In combination with the capability to swim vertically and horizontally, larvae of both invertebrates and vertebrates can orient and potentially navigate over short (meter-to-kilometer) to long (10-to-100-km) distances, using light, sound, smell, and possibly magnetism, electric fields, and wave swell (e.g., Kingsford et al., 2002;Gerlach et al., 2007 Knowledge of larval transport in nearshore environments is very limited. Major drawbacks include lack of rigorous knowledge of the suspected physical mechanisms involved in larval transport, and ignorance of other potential transport mechanisms (see Cowen, 2002, for a review). Physical mechanisms that could affect transport include surface gravity waves (Monismith and Fong, 2004), submeso-and mesoscale eddies (Bassin et al., 2005;Sponaugle et al., 2005), barotropic tidal currents (Hare et al., 2005;Queiroga et al., 2006), and cross-shore winds (Tapia et al., 2004).
Some proposed mechanisms have not been tested rigorously in field conditions. Moreover, the logistical difficulty of studying transport sometimes can push researchers to use weak inferential approaches, such as inferring larval transport mechanisms from settlement data (Pineda, 2000;Queiroga et al., in larger-scale models, thereby capturing the large-scale aspects of larval transport, the modulation of small-scale processes by large-scale forcing, and the very small-scale processes (e.g., turbulence) where larval swimming capabilities and behavior become overly important (see discussion in Metaxas, 2001). Even modeling a single, relatively straightforward process, such as the accumulation of particles in gravity currents, can be extremely complex (e.g., Scotti and Pineda, 2007). Thus, using numerical models for inferring larval transport when poorly studied processes may be important, or where the physical forcing is unknown, is dire. On the other hand, it is clear that numerical models are powerful tools in settings where processes are well known and in cases where field hydrodynamics are well simulated by the model (e.g., Reyns et al., 2006). Thus, we suggest that bottlenecks in understanding larval transport are less related to numerical modeling than to the mechanistic knowledge of larval transport.

Challenges of Adaptive Sampling
It is unclear how much larval transport occurs during episodic events and how much occurs during "mean" condi- Adaptive sampling, defined as sampling in response to an event, is a solution to these dilemmas; it has been used successfully to sample hydrodynamics and larval distributions during transport by internal tidal bores (Pineda, 1994(Pineda, , 1999. Adaptive sampling is challenging, however, because it is hypothesis based; sampling is initiated in response to a real-time change in a time-dependent variable, such as temperature or wind direction, that is integral to the hypothesized larval transport mechanism.
Adaptive sampling is therefore a stringent hypothesis test, because if larval transport does not occur as expected, the hypothesis is rejected. Adaptive sampling is also logistically difficult. If the events are sporadic, and the sampling is shipboard, adaptive sampling requires having a vessel and crew on standby ready to sample for long periods, an expensive prospect for anxious researchers.
Conceivably, remote sampling systems initiated in response to events could be constructed with off-the-shelf gear and new technologies currently under development such as in situ molecular detection of larvae (e.g., Goffredi et al., 2006). increasing and extending into nearshore areas (e.g., Chen et al., 2006). Decreased grid size, however, is only one aspect of resolving smaller-scale processes. Smallscale processes, such as surface waves, internal waves, and propagating convergences, need to be included. Currently, no numerical model appears capable of simultaneously resolving Lagrangian transport caused by, for example, shallowing internal tides, sea breeze, largeamplitude internal waves, and surface gravity waves. Further, accurately modeling larval transport will require embedding these small-scale processes First, hypotheses on the role of behavior in transport need to be developed and tested. Colby (1988) argued that passive advection and diffusion should be the null hypothesis for studies of larval transport.
In an early example of this approach, Woods and Hargis (1971) compared the distribution of coal particles with that of similarly sized oyster larvae and concluded that larvae were not being transported passively. A study on ascidian tadpole larvae found that dispersal distance was shorter in swimming larvae than in nonswimming individuals of similar size and shape (Bingham and Young, 1991). Similarly, Arnold et al. This approach should be expanded to take advantage of advances in modeling as well as in field and laboratory studies.
Behavioral hypotheses from laboratory studies are attractive because quantification of hydrodynamics and behavior is feasible, but these hypotheses should be tested in field conditions, and vice versa.
Second, the incorporation of behaviors into models of transport needs to be rule-based rather than deterministic, and individual variability should be considered. Most transport models that include larval behavior use population-level descriptions of distributions or swimming speeds and apply them to particles released in the model (Hare et al., 1999).
Another approach is to provide a set of behavioral rules that attempt to capture the trade-offs between feeding and predation; these rules result in vertical (and potentially horizontal) responses to various cues (Titelman and Fiksen, 2004;Fiksen et al., in press). Although the importance of time-dependent behaviors, such as diel, tidal, and ontogenetic, is well recognized, little is known about "adaptive" behavior on scales of seconds to minutes, where larvae might respond to transient physical and biological features. We know that larvae respond behaviorally to a number of factors, such as time of day, light, temperature, turbulence, pressure, and food availability, and that some of these responses influence transport, but only a few behaviors facilitating transport have been identified (e.g., Boehlert and Mundy, 1988;DiBacco et al., 2001). For example, field observations, modeling, and laboratory experiments imply that "swimming up" behaviors in response to transient downwelling flows in propagating features determine efficient larval transport (Pineda, 1999;Scotti and Pineda, 2007).
To incorporate our understanding of behavior into rule-based models will require a hypothesis-based approach. Without hypotheses, we run the risk of Third, most research has focused on how larval behavior affects advection, but the influence of behavior on diffusion requires more emphasis. Using an advection-diffusion-mortality model, Cowen et al. (2000) estimate that successful larval transport to coral reef habitats diminishes sharply when diffusion rates increase from 0 to 100 m 2 s -1 (the latter is a typical diffusion rate used in larval transport studies; see also Okubo, 1994). However, the assumption that larvae diffuse passively in the marine environment likely does not hold, particularly for older larval stages. Peaks in settlement must result from highdensity patches of larvae reaching adult habitats, and these coherent patches run counter to hypothesized diffusion. Natunewicz and Epifanio (2001) followed discrete patches of crab larvae for up to six days and hypothesized that associative swimming behaviors might be responsible for patch maintenance.
A U-shaped patchiness-at-age function has been described for the larval stages of several fish species, and this shape has been interpreted as initial diffusion with subsequent schooling (Matsuura and Hewitt, 1995). In addition, larvae may remain in thin layers of food (Lasker, 1975)

L ARVAL DISPER SAL Defining Dispersal Kernels
Most attempts to describe dispersal kernels have emphasized larval transport (e.g., Botsford et al., 1994), but other processes such as spawning, settlement, pelagic larval duration, and survival also influence larval dispersal (Edwards et al., in press). Many marine species release their offspring at specific locations and times, using specific behaviors. For example, relatively sedentary bluehead wrasse spawn daily at particular reef spawning sites that have been used for years (Warner, 1988). Similarly, several fish species spawn in circular motions that may create hydrodynamic vortexes (Okubo, 1988;Heyman et al., 2005). The influence of these small-scale events on larval dispersal over periods of weeks is unknown. On a larger scale, a number of motile species, including snappers, herring, and blue crabs, move to particular locations for spawning (Carr et al., 2004;Heyman et al., 2005). In the temporal domain, many coral species participate in annual mass spawning events, with more than 60% of species spawning over the course of several days (Babcock et al., 1994), and crabs and barnacles tend to release their larvae at certain phases of the tide or the day (Morgan, 1995;Macho et al., 2005). While such spawn-ing behaviors have long been thought to maximize larval survival (e.g., Hughes et al., 2000), the overall effect of localized and punctuated spawning on larval dispersal is unclear.  (Pechenik, 1986;Cowen, 1991). Some species have very narrow habitat requirements for the continuation of the life cycle, such as river mouths on isolated oceanic islands for some gobies, wave-beaten rocky points for gooseneck barnacles, and specific species of anemones for some reef fish (Radtke et al., 1988;Cruz, 2000;Jones et al., 2005). Other species have broad habitat requirements such as eurytopic Pachygrapsus crabs (Hiatt, 1948) and flounders of the genus Etropus (Walsh et al., 2006). For most species, only a subset of locations will support the continuation of the life cycle; these locations must be reached within the time window of possible settlement. Understanding these habitat and time constraints will be necessary to observe and model dispersal kernels. A number of models have included such considerations at a relatively large scale, for example, assuming modeled larvae that arrive within 10-15 km of known habitat have successfully settled (Hare et al., 1999;Paris et al., 2005). How larvae transverse these last 10 km is unknown largely because of the exclusion of smaller-scale processes in models and the inability to include realistic behaviors (see above).
The dispersal kernel also is dependent on larval mortality. Most studies of larval dispersal, however, either do not consider larval mortality (Hare et al., 1999), consider spatially homogenous mortality (Cowen et al., 2000), or assume low mortality (Gaylord and Gaines, 2000).
Modeling studies that assume low mortalities should be reconsidered in light of observed higher mortalities (e.g., Rumrill, 1990); use of high mortalities in dispersal models frequently yields lower maximum dispersal estimates than those obtained assuming low mortality (Cowen et al., 2000;Ellien et al., 2004;Tapia and Pineda, 2007). Differential survival of larvae during transport con- Larval duration also influences survival probability. Pelagic larval duration (PLD) must be correlated with the dispersal kernel for the simple reason that species with short PLD must have reduced larval transport and relatively "short" dispersal kernels; PLD is a constraining variable for dispersal. In contrast, long PLDs do not necessarily yield broad dispersal kernels, as larval behavior breaks the direct-proportional relationship between PLD and dispersal distance, both for fish and invertebrates (Sponaugle et al., 2002). Of course, long PLD yields higher cumulative mortalities than short PLD when everything else is equal (i.e., same daily mortality for species with short and long PLD; see Hare and Cowen, 1997). It is also unclear how variables influencing PLD, such as temperature and food (Scheltema and Williams, 1982), may influence the dispersal kernel (see O'Connor et al., 2007, for model predictions). Thus, the relationship between PLD and dispersal is ambiguous except for species with very short larval durations (see discussion in Sponaugle et al., 2002).

Dispersal Estimates in the Coastal Ocean
Given the complexity of larval dispersal, it is not surprising that measurement of a dispersal kernel in the marine environment is extraordinarily rare (Shanks et al., 2003). Gerrodette (1981) measured the dispersal of planula larvae from adults in a temperate solitary coral and found that mean dispersal distance from the parent was < 50 cm. Similar work with ascidians quantified dispersal from spawning to settlement, but the pelagic stage of ascidians is short (hours), larvae are large (millimeters), and mortality is low (< 90%) (Olson and McPherson, 1987), making it possible to follow individuals from the beginning to the end of the pelagic stage (see also Bingham and Young, 1991). Work on an isolated reef indicated that most acroporid and pocilloporid corals recruited in experimental moorings within 300 m from the reef, and that spat mortality decreased with distance from the reef (Sammarco and Andrews, 1989). Several studies followed patches of more typical marine larvae Oceanography Vol. 20, No. 3 32 Eventually, long-term, labor-intensive studies will be needed to increase our understanding of reproductive population connectivity of longer-lived mobile species.
(PLD of weeks, size < 1-10 mm, and high mortality), but these efforts are not true measures of larval dispersal because the spawning and ending locations were inferred (Pepin and Helbig, 1997;Natunewicz and Epifanio, 2001;Paris and Cowen, 2004). Other studies marked spawned eggs and then collected offspring at the end of their planktonic stage Almany et al., 2007); these studies provide a partial measure, but not a complete description, of the dispersal kernel because all potential ending locations could not be sampled.
Although dispersal kernels will eventually be fully quantified for some species in some systems, the measurement of these probability distributions in the marine environment will remain extremely rare.
It is easier to obtain dispersal kernels with models than with field measurements. Some models consider simpli-

Variation in Larval Traits and Survival During the Pelagic Stage
Most larvae exhibit variation in early life history (ELH) traits, such as size at a given age and growth rate. This variation can be introduced as early as the egg stage, when differential size, age, condition, or stress level of the mother can influence quality of the spawned eggs (Berkeley et al., 2004;McCormick, 2006 (Cushing, 1990;Baumann et al., 2006). Encounter with oceanographic features such as fronts or mesoscale eddies can also influence food supply and exposure to predators (Grimes and Kingsford, 1996;Sponaugle and Pinkard, 2004). Thus, a complex oceanographic environment coupled with variable egg quality at spawning results in a pool of larvae with variable traits (Jarrett, 2003;Lee et al., 2006;. Survival of pelagic larvae is typically nonrandom and proceeds according to three general concepts of the "growth-mortality hypotheses" (reviewed in Anderson, 1988). Theoretically, survivors should be those larvae that are larger at a given age ("bigger is better" hypothesis; Miller et al., 1988), grow faster ("growthrate" hypothesis; Bailey and Houde, 1989), and/or move through an early stage more rapidly ("stage-duration" hypothesis; Anderson, 1988 barnacles, bryozoans, and fishes (e.g., Searcy and Sponaugle, 2001;Pechenik et al., 2002;Jarrett, 2003;McCormick and Hoey, 2004;Phillips, 2004;. The potential exists for some traits that are advantageous to larvae to become subsequently detrimental to juveniles or vice versa. For example, crab zoeae reared at reduced salinities suffer higher mortality as larvae, but metamorphose into larger juveniles (Giménez and Anger, 2003), and a short pelagic larval duration enables fish larvae to escape the predation in the plankton, but results in smaller settlers (e.g., , which in some cases may be more susceptible to predation (Anderson, 1988 in an experimental manipulation, however, adults that were larger as larvae had higher survival rates and produced larger larvae themselves than those that were smaller as larvae, although delaying metamorphosis erased this relationship (Marshall and Keough, 2006). Optimal traits may vary with the environment encountered by the larval, juvenile, or adult stages, as evident for a snail (Moran and Emlet, 2001) and colonial ascidian . Thus, traits obtained during early stages have the potential for long-term effects on later stages, but many complex interrelation-ships likely influence the outcome. When carryover effects occur, they may persist, become amplified, or, instead, be compensated for during subsequent stages (Podolsky and Moran, 2006). In short, simply reaching a settlement site does not guarantee that larvae will possess the necessary traits to survive to reproduce.

POPUL ATION CONNECTIVITY: RE SE ARCH NEEDS
The for some purposes, is functionally insufficient. New efforts to track settlers to reproduction will initially advance with shorter-lived sessile species. Eventually, long-term, labor-intensive studies will be needed to increase our understanding …simply reaching a settlement site does not guarantee that larvae will possess the necessary traits to survive to reproduce.
of reproductive population connectivity of longer-lived mobile species. There is a rich history of marine ecological work examining the relative importance of recruitment versus density-dependent, post-settlement processes in structuring benthic populations (Caley et al., 1996), but we need to move beyond numerical responses and refine the question to focus on trait-based ecological linkages among all stages. Real measures of reproductive population connectivity require an understanding of who is surviving to reproduce and why.
As there is ample evidence that larval growth and condition can influence performance in later stages, from a practical point of view we need more reliable measures of condition. The coarsest measures of condition often use size as a proxy (e.g., many invertebrates), while others measure organic (Jarrett, 2003) or lipid content (Hentschel and Emlet, 2000), RNA/DNA ratios (Suthers et al., 1996;Lee et al., 2006), or (for fishes) otolith-based measures (e.g., , all of which have some limi- with physical-transport processes p and feeding and prey environments e, the vast parameter space that potentially affects pelagic eggs and larvae, and vexes researchers, may be effectively reduced to a more manageable set.