MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 567: 29–40, 2017https://doi.org/10.3354/meps12065

Published March 13

INTRODUCTION

Multiple factors contribute to setting the geo-graphic range of a species. Principal among these arethe interaction between the physiological tolerancesof the species, the distribution of environmental con-ditions, and, for some systems, interspecific interac-tions, especially competition (Hutchins 1947, Case &Taper 2000; a summary of the terrestrial literature isin the introduction of Eckhart et al. 2011). Dispersalcan modify how these interactions set the geographicrange of a species; in systems where dispersal is not

spatially biased (isotropic), it is primarily assumed toexpand range boundaries by providing subsidy fromregions of abundance, allowing the persistence of aspecies where it is less competitive (Gotelli 1991).

However, in many systems, dispersal is spatiallybiased (anisotropic). Spatially biased dispersal occursin marine and similar systems (e.g. rivers, streams,terrestrial systems with wind dispersal) when disper-sal is strongly influenced by currents or winds with apreferred directionality (Siegel et al. 2003). Gaylord& Gaines (2000) found that spatial variation in cur-rents could keep competitively neutral species sepa-

© Inter-Research 2017 · www.int-res.com*Corresponding author: jpringle@unh.edu

Ocean currents and competitive strength interact to cluster benthic species range boundaries

in the coastal ocean

James M. Pringle1,*, James E. Byers2, Ruoying He3, Paula Pappalardo2, John Wares4

1Ocean Process Analysis Group, Institute for the Study of Earth, Oceans, and Space, University of New Hampshire, Durham, NH 03824, USA

2Odum School of Ecology, University of Georgia, Athens, GA 30602, USA3Department of Marine Earth and Atmospheric Sciences, North Carolina State University, Raleigh, NC 27695, USA

4Department of Genetics, University of Georgia, Athens, GA 30602, USA

ABSTRACT: Dispersal of many coastal marine species is mediated by flows with strong direction-ality; bathymetric and topographic effects lead to strong alongshore variability in this transport.Using a simple model of the population dynamics of competing benthic species in a coastal ocean,we found that alongshore variability in dispersal can lead to clustering of species range bound-aries for species whose dispersal is dominated by coastal currents. Furthermore, species can beabsent from areas where they would have a relative competitive advantage because the presenceor absence of a species is determined not only by local conditions but also by propagule supply,which is often affected by larval transport from far upstream. Our model demonstrates the quanti -tative linkages between alongshore variation in coastal currents, spatial gradients in competitivestrength, and the geographic extent of a species. We show that the predictions of the model areconsistent with observed species distributions in the Gulf of Maine and Mid-Atlantic Bight, USA.A mechanism for extensive coexistence of competing species where range boundaries cluster isdescribed. The implication of the clustering highlighted by our model suggests that for specieswhose dispersal is dominated by long-distance planktonic periods, climate change inducedchanges in the relative competitiveness of species will lead to abrupt changes in species rangeboundaries and not gradual range extension.

KEY WORDS: Range limits · Drift paradox · Biogeography · Dispersal · Advection · Larvae

Resale or republication not permitted without written consent of the publisher

Mar Ecol Prog Ser 567: 29–40, 2017

rate. Byers & Pringle (2006) found that in the absenceof inter-species competition, anisotropic dispersalcan reduce a species range, shifting it downstreamand increasing the fecundity necessary for a speciesto persist at any location. They emphasized thatwithin marine systems, there are fluctuations in theflow driven by eddies or temporal variation in thewinds that allow larvae to move ‘upstream’ againstthe mean currents, and emphasized the importanceof these ‘upstream’ transports to population persist-ence. Here we build on the work of Byers & Pringle(2006) to include effects of interspecies competition.

We examined how environmentally driven gradi-ents in relative competitiveness or fecundity of a spe-cies interact with alongshore variation in larval trans-port and retention to establish the upstream-mostextent of its range. The result is a criterion for themaintenance of the upstream edge of a geographicrange as a function of local circulation and the rela-tive competiveness of 2 competing species. This workextends previous studies by explicitly modeling theinteraction between environmentally driven gradi-ents in relative competitiveness or absolute fecundityand realistic downstream-biased dispersal (Gaylord& Gaines 2000 did not include competition; Case etal. 2005 only considered isotropic diffusion). The re -sults will help us to understand what fixes speciesranges in marine and other systems with down-stream-biased dispersal, where competing specieswith biased dispersal can coexist, and how their dis-tributions will respond to changes in climate.

In coastal oceans, clusters of species boundariesare often found to occur in regions where there arestrong spatial gradients in water temperature. Theseclusters have been attributed to temperature toler-ances and to the dependence of relative competive-ness on temperature, and to the increased likelihoodthat the environmental limit of a species will bereached in a location where those environmentalconditions are changing rapidly in space (Hutchins1947). In explicit tests of the relative importance ofenvironmental gradients (like temperature) and dis-persal in fixing species ranges, neither alone receivesoverwhelming support (summarized by Gaines et al.2009). We examined how alongshore variation inalongshore dispersal can create clusters of rangeboundaries on their own, and how the same circula-tion patterns that cause this variation will also lead toregions of enhanced temperature gradients. This willclarify how alongshore variability in currents cancontribute to range boundaries directly, by alteringdispersal, and indirectly, by altering the alongshoredistribution of temperature.

As a test case, we used circulation data derivedfrom surface drifters for the Gulf of Maine and theadjacent part of the Mid-Atlantic Bight, USA (Man-ning et al. 2009), to predict the location of clustersof upstream species boundaries. These predictionswere compared to observed biogeographic patternsin benthic marine invertebrates. Benthic marineinvertebrates, a group that dominates marine macro-scopic biodiversity, provide an ideal test group inwhich to examine the interactions of environmentalgradients and circulation. The adults are often sessileor have limited dispersal, and their dispersal is domi-nated by a larval stage. Available data encompassmany phyla (e.g. Arthropoda, Mollusca, Echino -dermata, Annelida) that contain many species with aplanktonic larval stage in their life cycle for whichdispersal is mediated by ocean currents (Strathmann1987). Both the circulation and species distributionsin this region are well studied, and the area containsseveral well-established biogeographic boundaries(e.g. Bay of Fundy, Cape Cod).

METHODS

Criterion for existence of upstream boundaries

Given complete knowledge of the population con-nectivity, habitat suitability, and relative competi-tiveness of 2 benthic species with planktonic disper-sal, it would be straightforward to describe how theywould interact to set each other’s ranges. However,this knowledge is rarely straightforward to obtain(Connolly & Roughgarden 1999). Instead, we exam-ined the statistics of larval delivery to a location todiscover if that point could be the upstream-mostrange boundary for a given species in an advectiveenvironment. We focused on the upstream edge be -cause once a species can persist at a point, its per -sistence farther downstream can be partly (andsometimes largely) driven by larval subsidies fromupstream. Thus criteria are found for a local condi-tion that allows the species to persist on a regionalscale. We assumed that while environmental toler-ances are ultimately defined by the limited adapt-ability of species, sensu Eckhart et al. (2011) and Sexton et al. (2009), the locations of their rangeboundaries within the region they can tolerate can bemodified by dispersal and competition.

Two cases are considered: (1) two interacting spe-cies that have the same dispersal and produce thesame number of larvae that eventually reach habitat,but which compete for habitat and have spatially

30

Pringle et al.: Currents cluster coastal marine range boundaries

varying and different abilities to compete for thehabitat. This case examines differences in relative fit-ness caused by post-dispersal processes includingrelative abilities in recruitment and growth com-pared to another species with which it sharesresources (especially space and food); (2) two inter-acting species that compete equally for habitat oncelarvae reach suitable habitat but have different num-bers of larvae which reach the habitat. This caseexamines differences in relative fitness caused bypre-settlement processes, including spatially differ-ing production of larvae and spatial differences inlarval mortality while in the plankton.

Thus, our 2 cases cover dynamics either pre-settle-ment or during and after settlement in systems withcompetition. The 2 cases are equivalent when disper-sal is spatially unbiased (Sexton et al. 2009 and cita-tions therein). However it is not clear that this shouldbe so when currents spatially bias dispersal (e.g.anisotropic dispersal), because the effect of inter-species differences pre-dispersal might be shifteddownstream by the spatially biased dispersal. Foreach case, a criterion will be developed for the per-sistence of the downstream species at and down-stream of its upstream limit.

Case 1: Two species that differ during and aftersettlement

We assume a competitive interaction between 2species such that the total population of the 2 speciesis set by the carrying capacity of the habitat, but therelative ability of each species to compete for thehabitat is a function of location. The ‘upstream’ spe-cies is favored upstream of a location, and the ‘down-stream’ species is favored downstream of that loca-tion. We also assume that individuals of both speciesproduce the same number of larvae that survive dis-persal and that could recruit and reproduce in theabsence of density dependence and competition. Acriterion is found for the minimum inter-species dif-ference in competiveness at the upstream rangeboundary (‘the boundary’) of the downstream speciesto allow the downstream species to persist at thatlocation. The downstream species does not existupstream of the boundary.

The criterion is based on a single robust assumptionabout conditions at the boundary of the downstreamspecies: for the downstream species to persist at theboundary, the fraction of its larvae that recruit to thatlocation and survive to reproduce there (relative to allcompeting species’ larvae that reach that location)

must be equal to the fraction of adults of the down-stream species relative to competing adults of theother species at the boundary. This criterion exists be-cause the relative frequency of the 2 species in thenext generation at a location is set by the fraction oflarvae that settle and survive to reproduce there. Ifthe fraction recruiting and surviving to reproduce isless, the species fails to replace itself and the fractionof the downstream species at that point will continu-ally decrease over generations until the point of localextinction. In the development below, generations arenon-overlapping; the results are robust to deviationsfrom this assumption. Overlapping generations wereexamined by Pringle et al. (2009), who found that per-sistence of the downstream species depends on thenumber of larvae that recruit and survive to reproduc-tion over the adult lifetime. Thus the statementsabove apply when averaged over an adult’s lifetime,weighted by recruitment in each reproductive event.

We parameterize larval delivery to the upstreamboundary with 2 parameters: ƒret, the fraction of allcompeting larvae (of either species) returning to theboundary that originate at and downstream of theboundary, and ƒup, the fraction that come fromupstream of the boundary (and must, by definition,only include larvae of the upstream species). ƒret + ƒup

= 1, and these parameters can be estimated as belowfrom knowledge of circulation (see the Supplementat www.int-res.com/articles/suppl/ m567 p029_ supp.pdf [Section SI-1]), or could potentially be estimatedfrom observations of larval settlement at the bound-ary (e.g. DiBacco & Levin 2000). We make a furtherlimiting assumption that the dispersal parameters arethe same for each species; this assumption is relaxedin the Discussion and in the Supplement (see SectionSI-5) (cf. Lutscher et al. 2007, Salomon et al. 2010,Bode et al. 2011, Aiken & Navarrete 2014).

The larval delivery parameters ƒret and ƒup are thenused to understand the relation between the relativefraction of adults of different species at the boundaryto the relative frequencies of the larvae of the differ-ent species arriving at the boundary (Fig. 1). At theboundary, the fraction of adults of the upstream anddownstream species are Pup and Pdown, so Pup + Pdown

= 1. The fraction of larvae reaching the boundarylocation (though not necessarily recruiting and sur-viving to reproduction) is Lup for the larvae of theupstream species, and Ldown for those of the down-stream species (Lup + Ldown = 1). As assumed above,each species produces the same number of larvae peradult which survive dispersal. Lup is then Pupƒret + ƒup.Pupƒret is the product of the fraction of the upstreamspecies in the downstream region, Pup, with the frac-

31

Mar Ecol Prog Ser 567: 29–40, 2017

tion of the larvae from that region, ƒret, and ƒup is theproduct of the fraction of the upstream species in theupstream region (1.0) and the fraction of the larvaethat arrive to the boundary from upstream, ƒup. Like-wise, the relative fraction of larvae of the down-stream species at the boundary is Ldown = Pdownƒret.

In a case where there is no difference in competi-tiveness between the upstream and downstream spe-cies, so each larvae has an equal chance to success-fully recruit, the fraction of the downstream speciesin the next generation ( ) will be proportional tothe fraction of the downstream type in the larvaereturning to the boundary (using Pup + Pdown = 1):

(1)Since ƒret must be <1 if there is any larval inputfrom upstream, Pdown must decrease each generation.

Thus without competitive differences, there can beno stable boundary between the 2 species if there isany larval input from upstream. As an example, inFig. 1 we see that the fraction of the downstream spe-cies downstream of the boundary is 2/3, but only20% of all larvae arriving at the boundary are of thedownstream type (Ldown). This is because many of thelarvae arriving at the boundary are from upstream(ƒup > 0) and of the upstream species and becausesome of the larvae arriving to the boundary fromdownstream are of the upstream type (Pup > 0). In theabsence of any competitive advantage for the down-stream type, this low return of downstream larvaewould reduce the frequency of the downstream species to 20% over an adult lifetime; this frequencywould continue to decline with time until local ex -tinction of the downstream species.

However, if ΔC is the fractional increase in relativecompetitiveness, and a single larva of the down-stream species is (1 + ΔC ) more likely to successfullyrecruit and reach reproductive competency down-stream of the boundary than a larva of the upstreamspecies, we can re-write Eq. (1) as

(2)

In this case, there can be a stable boundary. To findit, we set Pdown

next to Pdown in Eq. (2) to find a ΔC forwhich Pdown does not change in time. Then by solvingfor ΔC in the limit that Pdown approaches 0, we canfind the minimum ΔC required for a boundary to persist:

(3)

Case 2: Two species that differ pre-settlement

Under this scenario, the species differ in the num-ber of larvae that reach a habitat, but each larva isequally competitive for that habitat. The relative dif-ference in the number of larvae produced per adultthat can reach a given site is ΔR, such that the down-stream species has (1 + ΔR) greater larval deliveryper adult than the upstream species at the boundary.ΔR is chosen as the notation because one source ofthe difference in larval number could be spatiallyvarying fecundity, but the difference could also bedriven by differing mortality during dispersal. Wecan then write

Pdownnext

ƒƒ ƒ ƒ

ƒdownnext down

down up

down ret

down ret up ret updown retP

LL L

PP P

P=+

=+ +

=

PC L

C L L

C P

downnext down

down up

d

))

)

= ++ +

= +

((

(

11

1

ΔΔ

Δ oown ret

down ret up ret up

ƒ) ƒ ƒ ƒ(1+ + +ΔC P P

ƒ

ƒ1 ƒ

ƒmin

up

ret

ret

retCΔ = = −

32

Fig. 1. Model and hypothetical numerical example of themaintenance of the upstream boundary (dashed line) of aspecies (green circles) against the effects of a mean current(black arrow) that transports most larvae downstream. Thedownstream species does not exist any farther upstreamthan this upstream boundary. The upstream species is rep-resented by the red triangles. ƒup: fraction of all larvae arriv-ing to the boundary that were produced upstream of it, ƒret:fraction of all larvae arriving to the boundary that were pro-duced at the boundary region as well as locations fartherdownstream. Pup: fraction of adults at and immediatelydownstream of the boundary composed of the upstream spe-cies, Pdown: fraction of adults immediately downstream of theboundary that are of the downstream type. From these pa-rameters, we can compute the fraction of larvae arriving atthe boundary which consist of the upstream species (Lup)and the downstream species (Ldown). The relative size ofthese larval deliveries is indicated by the thickness of the ar-rows, and the arrows are labeled with the term of the equa-tions for Lup and Ldown that each arrow represents (red: up-

stream species, green: downstream species)

Pringle et al.: Currents cluster coastal marine range boundaries

and (4)

where Ldown and Lup are normalized so that Ldown +Lup = ƒret + ƒup = 1, as above. Inter- and intraspecificcompetition leading to density dependence is as -sumed to be of the same intensity (sensu Case et al.2005). The role of ΔR in aiding retention of the down-stream species in the next generation can be ex -pressed as:

(5)

which can be solved as above for the minimum ΔRrequired for a boundary to persist:

(6)

This result is identical to the previous case (Eq. 3),illustrating that, as in systems with isotropic disper-sal, a change in either relative pre-settlement larvalproduction and survival or relative post-settlementcompetitiveness have the same effect on persistenceof an upstream species boundary (cf. Sexton et al.(2009) for similar results with isotropic dispersal).Hereafter, we shall treat these 2 cases as equivalent,and thus when we discuss the effects of spatial varia-tion in competitiveness, it should be read to alsoapply to spatial changes in the relative fecundity/mortality of 2 species.

Population structure downstream of the rangeboundary

The criterion for the persistence of a downstreamspecies at the boundary does not mean that a com-petitively inferior upstream species must have adownstream boundary at the same point. The up -stream species, because it is supported by a larvalsubsidy from upstream, will exist for some distancedownstream of the boundary before being van-quished by the competitively superior downstreamspecies. This can be illustrated by re-writing Eq. (2)to be applicable downstream of the boundary (soPdown can be non-zero at farther upstream points) andin steady state (so Pdown is the same from generationto generation). In these limits, where P n

down is the population of the downstream species at location n

and n increases downstream, and and Pdown can besolved iteratively from the upstream boundary with

(7)

using Pdown + Pup = 1 to solve for Pup. These solutionswill be described further in the discussion.

Interpretation of criteria

If a species can persist at a point, the species willalso persist downstream of the point where it has arelative competitive or fecundity advantage becauseloss of larvae downstream can be balanced by larvalimmigration from upstream. This can easily be seenfrom Eq. (7), where if the population of the down-stream species exists at a point (n −1) where it isfavored, it must also exist at the next downstreampoint (n) as long as there is some larval input fromupstream (i.e. ƒup is not 0). Thus, the local condition inEq. (3) for the persistence of an upstream-most species boundary is a regional condition for the persistence of a species where the time-averagedcurrents continue in the same direction (i.e. ‘down -stream’ does not change direction).

Eq. (3) shows that as the fraction of larvae retainedin a region ƒret increases (as one would expect forshort larval duration), the relative competitivenessadvantage needed for a species to have an upstreamboundary at a location limits to 0. In this limit, thelocation of a species range boundary becomes con-trolled by where the relative fitness of competingorganisms changes from favoring one to the other(ΔC changes from positive to negative, passingthrough 0). Thus, in the limit of very short larvalduration and thus small dispersal, persistence is gov-erned by local differences in relative competitivenessand the nature of dispersal becomes less important.This is similar to the results with isotropic dispersal(for example, for many terrestrial species), where thetransition be tween ranges of competing species is,in the absence of other mechanisms to bias rangeboundaries, centered on where the relative competi-tiveness of the competing species changes fromfavoring one to the other species (Sexton et al. 2009).

There are subtleties that are concealed in the deri-vation of the criteria for Eq. (3). It is likely that the rel-ative competitiveness between 2 species is a continu-

= + −ƒ ƒup up ret up1

upL P Pn n n

= + −ƒ ƒdown down ret down1

upL P Pn n n

PC L

C L L

C P

ndown

down

down up

up

))

ƒ

= ++ +

=

+ ⋅

((

( )

11

1

ΔΔ

Δ ddown ret down

up up ret up

ƒ

ƒ ƒ

n n

n n

P

P P C

−

−

+( )+ + +

1

1 1( )Δ ⋅⋅ +( )−ƒ ƒup down ret downP Pn n1

(1 ) ƒ(1 )down

next down

down up

down ret

down upP

LL L

R PR P P

=+

= + Δ+ Δ +

ƒ

ƒ1 ƒ

ƒmin

up

ret

ret

retRΔ = = −

(1 ) ƒ(1 )

downdown ret

down upL

R PR P P

= + Δ+ Δ +

ƒ

(1 )ƒup

up ret

down upupL

P

R P P=

+ Δ ++

33

Mar Ecol Prog Ser 567: 29–40, 2017

ous function of environmental condi-tions and thus a continuous functionof location. However, in Eq. (3), were present these parameters at the up -stream boundary of a species rangewith a single competitive value. Weshould think of this single value asthe average for those adults whoselarvae are returned to the boundary.Thus ΔC is an average of the compet-itiveness difference over a spatialextent defined by the dispersal of lar-vae to the upstream boundary. In theSupplement (Sections SI-2 and SI-3),we show how this spatial extent canbe quantified as a function of the ob -served larval dispersal for coastlineswith simple geometries.

Nonetheless, from the point of viewof the downstream species, an in -crease in larval retention (ƒret) decrea -ses the competitiveness advantage orfecundity needed for its upstreamrange boundary to persist, while anincrease in larval input from theupstream species (ƒup) increases thecompetitive advantage needed for theboundary to persist. More generallythan Gaylord & Gaines (2000), whoused idealized coastal circulations and neutral spe-cies interactions, we find that even subtle alongshorevariations in alongshore transport that increase thelocal retention of larvae or reduce larval input fromup stream will increase the likelihood of a speciesboundary persisting.

Oceanographic data

To qualitatively test these criteria against observedspecies distributions, we must estimate larval dis -persal and retention. This was done for the ScotianShelf/Gulf of Maine region. The mean currents(Fig. 2) from the Scotian Shelf to Cape Hatteras are tothe southwest, largely parallel to the coast (Lentz2008). At Cape Cod, the coastal current di verges off-shore, leading to a region of sluggish circulation inthe Nantucket Shoals region (Manning et al. 2009).The Bay of Fundy is a well-known region of en -hanced retention (Aretxabaleta et al. 2008, 2009).

Connectivity was defined by 911 NOAA driftersreleased from 1998 through 2013 from the Gulf ofMaine to the Mid-Atlantic Bight (Manning et al.

2009). Most drifters were drogued to 1 m depth andreleased in the spring or summer. The drifter tracksare shown as a function of region in Fig. S1 in theSupplement. A drifter is counted if it enters a regionand is still functioning a time T later; if at that time itis in one of the regions, the 2 regions are connected.T is a proxy for the larval duration; be cause thedrifters are at the surface where the currents aregreatest and are without behavior, this is an upperlimit of dispersal (Lentz 2008). A single drifter maydefine connectivity between multiple regions. Not alldrifters that survive a time T connect regions; somedrift out of any region, either downstream or off-shore. These data are used to define a connectivitymatrix E such that each element

Eij = nij/Ni (8)

where nij is the number of drifters that leave region iand a time T later are in region j and Ni is the numberof drifters that leave region i and are still functioninga time T later. The resulting connectivity matrix isused to calculate ƒret (Fig. 3) for the regions shown inFig. 2 and for a time in plankton T of 3 and 15 d.These planktonic durations are typical for short- and

34

Fig. 2. Surface coastal currents (thinner lines + arrows) of the Gulf of Maine,bathymetry, and modeling regions (thick lines). The shaded regions are mar-ine eco-regions where there is an enhanced number of northern rangeboundaries for species with long larval duration (Spalding et al. 2007, Pap-

palardo et al. 2015). The 100, 200, 1000, and 2000 m isobaths are shown

long-distance planktonic dispersers in the Gulf ofMaine (Pappalardo et al. 2015). In the Supplement(Section SI-4), E is used to parameterize an inter -active population model illustrating the interaction of2 competing species for a wider range of dispersaltimes (Fig. S4).

The alongshore temperature is used to compare al-ternative theories of range boundary maintenance be-low. The seasonal mean temperature from the WorldOcean Atlas 2013 is extracted from points along thecoast in the study region for the oceanographic sum-mer (July, August, and September) and winter (Janu-ary, February, and March) (Locarnini et al. 2013).

RESULTS AND DISCUSSION

The criteria developed above predict that in com-peting species, the downstream species can maintainits upstream boundary with a reduced competitive-ness differential (ΔC ) where ƒret, the fraction of larvaerecruiting that originated locally, is larger. Thus, allother things being equal, we would expect upstreamspecies boundaries to cluster in regions of high larvalretention (where ƒret is a local maximum). This pre-diction can be compared to the observations of Pap-palardo et al. (2015) of enhanced numbers of up -stream range boundaries in the Bay of Fundy and atCape Cod for benthic marine organisms whose meandepth of occurrence is less than 20 m. These ob -served boundary clusters are asymmetric; they in -clude many upstream limits for species, and fewerdownstream limits.

In the analyses of longer larval duration (T = 15 d,Fig. 3B, longer durations in Fig. S4), there are clearlocal maxima in ƒret at the Cape Cod/NantucketShoals region, and in the Bay of Fundy. At Cape Cod,the coastal current diverges offshore, leading to aregion of sluggish circulation and enhanced tidalmixing in the Nantucket Shoals region and reducedconnectivity with upstream regions (Manning et al.2009 and Fig. S1), and the Bay of Fundy is a well-known region of enhanced retention (Aretxabaleta etal. 2008, 2009). This is consistent with the observa-tions of clusters of range boundaries for species withlong larval duration (and thus presumably potentialfor long larval dispersal distances) in Pappalardo etal. (2015). These dynamics can be seen in the interac-tive population model presented in Fig. S4 at www.int-res. com/ articles/ suppl/ m567 p029_ supp.html, where3 variables can be manipulated: (1) the location and(2) magnitude of the change in competitive strength,and (3) the larval duration of the competing species.As predicted above, the location of an upstreamboundary for competing species with long larvaldurations does not directly follow the location of thechange in relative competitiveness. Instead, theboundary moves to the next downstream region ofrelatively high retention (the enhanced ƒret at CapeCod and the Bay of Fundy) that is most nearly down-stream of the point in space where a species starts tobe competitively dominant again.

In the analyses of short larval duration (T = 3 d,Fig. 3A), there are no clear isolated maxima in larvalretention ƒret at either Cape Cod or the Bay of Fundy,or elsewhere along the coast. This suggests that the

Pringle et al.: Currents cluster coastal marine range boundaries 35

Fig. 3. ƒret, the fraction of larvae recruiting that originated locally and downstream relative to total recruitment, for a larval du-ration of (A) 3 d and (B) 15 d. A high proportion of red indicates a retentive location. The gray lines separate the marine eco- regions as shown in Fig. 2. No value for ƒret is shown for the Cape Sable Island region because ƒret cannot be calculated for theupstream-most region, since the data do not allow us to estimate the fraction of larvae settling there from even farther

upstream. All fractions are estimated from surface drifter data

Mar Ecol Prog Ser 567: 29–40, 2017

previously observed peaks in upstream boundariesfor species with short larval duration in both regions(Pappalardo et al. 2015) is not driven by alongshorevariation in retention. However, it is worth notingthat at Cape Cod, the fraction of upstream bound-aries is larger for species with longer larval durationsthan shorter, suggesting that the long-distance dis-persers are more sensitive to this oceanographic feature (Pappalardo et al. 2015).

Clusters of range boundaries have also beenexplained by regions of increased gradients in waterproperties such as temperature. Essentially, theboundaries of a critical tolerance window for a givenenvironmental variable is more likely to exist within

a region if the water properties change greatly withinthat region (Hutchins 1947). However, in the Gulf ofMaine, temperature gradients can offer at best a par-tial explanation. In Fig. 4, the coastal temperature inwinter and summer is presented. At Cape Cod, thereis a localized dip in temperature at the cape thatbriefly interrupts a regional poleward cooling trendin both winter and summer, but it is limited to a25 km stretch of shoreline and is associated withenhanced mixing in Nantucket Shoals (Limeburner& Beardsley 1982). It seems unlikely that a regionalfeature of such limited extent would set rangeboundaries for species with long larval durations, asit is smaller than the dispersal distance of these lar-

vae (e.g. see dispersal estimates fromSiegel et al. 2003, Pringle et al. 2011, andFig. S1). We would expect larvae with longplanktonic durations to be able to passacross it in both directions in each genera-tion. Thus it is unlikely that the clusterof range boundaries at Cape Cod is ex -plained by this localized anomaly in sur-face water temperature. The situation issomewhat different near the Bay of Fundy.Immediately poleward of the bay, near thesouthern-most extent of Nova Scotia, thereis a sharp poleward drop in temperature inthe winter, and a sharp poleward increasein temperature in the summer. These tem-perature transitions are associated withen hanced mixing in the bay and diver-gence of a portion of the alongshore flowwhere the Scotian Shelf enters the Gulf ofMaine, and extend over the entire ScotianShelf (Hannah et al. 2001, Aretxabaleta etal. 2008). While the seasonal reversal ofthe temperature change is unusual, onecan create plausible scenarios for howit would anchor the upstream speciesboundaries (e.g. interacting sets of specieswhose relative fitness is more sensitive towinter than summer temperatures).

Nonetheless, the work above does notsuggest that the regional alongshore gra-dients in temperature (or other water prop-erties) are unimportant. The criteriongiven above in Eq. (3) requires a differ-ence in competitiveness for an upstreamspecies boundary to persist, and theremust be some spatial variation in the dif-ference in competitiveness for one speciesnot to be more competitive everywhere(and thus to exist everywhere). Competi-

36

Fig. 4. (A) Summer surface temperatures (°C) from the World Ocean Atlas 0.25° dataset. Red stars mark locations labeled in panel B. (B) Sea-sonal average coastal surface temperatures along the coast for both

winter and summer

Pringle et al.: Currents cluster coastal marine range boundaries

tion is highly context dependent, and the spatial gra-dient in competitiveness could well be due to gradi-ents in temperature, salinity, or other water proper-ties that affect competitiveness.

Further, in most coastal oceans, there would be astrong correlation between enhanced larval reten-tion and enhanced regional gradients in tempera-ture. The Gulf of Maine is unusual in that much of itstemperature structure is caused by regional variationin intense tidal mixing (Robinson & Brink 2006). Inmore typical coastal oceans, the alongshore variationin temperature is controlled in part by alongshorecurrents. This correlation is illustrated by a simplemodel of alongshore temperature variation whichhas proven successful in the Mid-Atlantic Bight, andwhich should be informative along many coasts inthe absence of intense upwelling or tidal mixing(Lentz 2010). The model assumes that the tempera-ture evolution of a parcel of water at the surface issimply the surface heat flux into the parcel, and thealongshore variation of temperature is given by theintegral of the surface heat flux along the path of sur-face water. A necessary result of this model is that thealongshore gradient in temperature is proportional to(alongshore velocity)−1, all other things being equal.This relationship occurs because the temperature ofthe parcel is changing at a fixed rate per time due to

surface heating or cooling — thus, the more slowlythe parcel moves, the greater the spatial gradient intemperature. Therefore, anything that interrupts orreduces the alongshore transport of water parcels(simultaneously reducing the alongshore velocity ofthat parcel and thus increasing ƒret) will increase thealongshore temperature gradient. This suggests wewould often expect the alongshore variations inretention and temperature to interact to concentratethe upstream range limits of many species at thesame locations.

The species range dynamics described above alsosuggest a mechanism for the coexistence of specieswith different relative competitiveness. The popula-tion of the upstream species can remain substantialeven well downstream of the point where a competi-tively superior downstream species can persist. Thisis illustrated in Fig. 5, where solutions to Eq. (7) areshown for an ƒret everywhere equal to 0.5 and a rela-tive competiveness (ΔC ) of 5, 50, and 100% greaterthan needed for the downstream species to persist;the upstream species remains present even wherethe downstream species exists. Thus, downstream-biased dispersal shifts the boundary of an upstreamspecies downstream of where it ceases to be compet-itively superior; its abundance slowly decreases withincreasing distance downstream of the transition.

These results are consistent with thespecies boundaries documented byPappa lardo et al. (2015), where up -stream species boundaries are moreclustered than downstream bound-aries. The ability of dispersal to main-tain a species where it is competi-tively inferior has been described inthe ecological literature (e.g. Gotelli1991 and citations therein); what ourwork illustrates is that these dynam-ics will be ubiquitous in the coastalocean due to the asymmetric dis -persal of larvae by ocean currents(see Nagylaki 1978), Pringle et al.2011, and Pringle & Wares 2007 forsimilar ideas in a population-geneticscontext).

Several other dispersal-driven me -ch anisms for co existence of specieswith differing competiveness havebeen described. Those of Lutscher etal. (2007), Salomon et al. (2010), andBode et al. (2011) differ from our workbecause they depend on the differ-ence in the dispersal distance of the 2

37

Fig. 5. Spatial evolution of the species favored upstream (red stars) and down-stream (blue squares) of location 1, where ƒret, the fraction of larvae recruitingthat originated locally and downstream relative to total recruitment, is 0.5everywhere, and ΔC, the change in relative competiveness, is 5, 50, and100% greater than the minimum ΔC, which allows the downstream species

to persist

Mar Ecol Prog Ser 567: 29–40, 2017

competing species. Our mechanism differs fromthose described by Aiken & Navarrete (2014) andBerkley et al. (2010) because their mechanismsexplicitly neglect source/ sink dynamics and aredriven by stochastic circulation and populationdynamics.

Efforts to estimate environmental tolerances of aspecies from its distribution will be confounded bylarval dispersal. As described above, a species whosedispersal is long and determined by local currentscan be absent where it is competitively superior andpresent where it is competitively inferior; in bothcases, its range is shifted downstream. Thus efforts toestimate its environmental tolerances will be biasedtowards conditions prevalent downstream of wherethe species is truly best adapted wherever the meancirculation biases larval transport downstream. Thiscan potentially confound efforts to use ecologicalniche models to estimate climate-induced shifts inspecies ranges (Phillips et al. 2006). These ideas havealso been presaged by similar analyses in the popu-lation-genetics context (Hare et al. 2005). Likewise,the interactions of 2 species at a spot on the coastlinecannot be fully understood without knowledge ofthe overall distributions of their ranges and their dis-persal. Coexistence (or the lack thereof) may havenothing to do with a site-specific measure of competi -tive strength, but may be due to immigration fromelsewhere.

The pinning of upstream range boundaries to re -gions of anomalous alongshore larval transport willalter how benthic species with long planktonic dis-persal stages will respond to climate change. It hasbeen assumed that species with short-distance dis-persal will respond more slowly to changes in climate(Harley et al. 2006); it is assumed, often tacitly, thatchanges in range boundaries are only limited by theirability to disperse to newly congenial habitat. How-ever, in coastal environments with mean currents,the short larval duration species can track changes inthe environment more closely than species with long-distance larval dispersal, as long as their upstreaminvasion speed is greater than the local climatevelocity (sensu Pinsky et al. 2013). Species withlonger larval dispersal may remain pinned to re -gional oceanographic features that modify connec-tivity, and thus fret, until they are favored in a similaroceanographic feature farther upstream. Because theretentive circulation features in the Bay of Fundy,Nantucket Shoals/Cape Cod, and elsewhere (e.g.Mace & Morgan 2006, Vander Woude et al. 2006) areoften bathymetrically driven, their location shouldnot change much in a changing climate. Thus, we

expect to see range expansion in species with longlarval durations as a mixture of periods of stasispunctuated by abrupt upstream movement of rangeboundaries to the next retention region, while spe-cies with short larval durations may change theirranges more continuously along the coast. Theseresults are an interesting complement to results sug-gesting that species with the ability to disperse byswimming against the mean currents (e.g. fish) canclosely track changing environmental conditions(Pinsky et al. 2013, Sunday et al. 2015).

It is worth noting that the models of competing spe-cies developed above can also be applied to alleleswith different fitness within a single species — e.g.see the similarity between the species level modeland results of Byers & Pringle (2006) and the evalua-tion of intraspecific diversity of Pringle & Wares(2007). This concordance could go a long way to ex -plaining the overlap of regions of many biogeo-graphic transitions and regions of many allelic clines(Wares 2002, Pelc et al. 2009, Altman et al. 2013,Ewers-Saucedo et al. 2016). The coexistence mecha-nism described above may also suggest an explana-tion for why marine protected regions set aside toprotect high species diversity also often serendipi-tously protect regions of enhanced genetic diversity(Small & Wares 2010).

While it qualitatively fits the distribution of bound-aries in the study region, the model of species bound-aries in competing species developed above is farfrom a complete description of how ocean transportinteracts with spatially varying inter-species inter -actions to set species boundaries. To gain a deeperunderstanding of how dispersal structures rangeboundaries in the ocean, one could begin by trying tounderstand how species with different dispersalstrategies interact (e.g. building on Lutscher et al.2007, Salomon et al. 2010, Bode et al. 2011, andAiken & Navarrete 2014; see discussion in the Sup-plement, Section SI-5). Furthermore, our model onlyincludes simple competitive interactions between 2species. The focus on competition is certainly justifi-able since it is a major process governing communityassembly rules. Many theories and models have beenbuilt on the colonization−competition tradeoff thatis manifested in species life histories (e.g. lotterymodel, supply-side ecology, etc.), especially for habi-tats like the rocky intertidal and shallow subtidalzones where space is often a limiting resource. However, other interactions, for example mutualisticinter actions and predation, may have a role in settingcommunity structure, but their interactions with dis-persal in marine systems has not been well studied.

38

Pringle et al.: Currents cluster coastal marine range boundaries

While we are confident that the dynamics describedhere will be important in many coastal oceans, wethink much more remains to be understood.

CONCLUSION

Prior efforts to quantify the role of coastal ocean cir-culation in structuring biogeographic or phylogeo-graphic boundaries in the ocean have either assumedhighly idealized coastal circulation (Gaylord &Gaines 2000) or highly idealized descriptions of lar-val dispersal (Byers & Pringle 2006, Pringle & Wares2007), or they have analyzed an entire connectivitymatrix (e.g. Cowen et al. 2006). The former ap -proaches are so idealized they can be hard to applyin typical coastal settings; the latter requires a daunt-ing amount of larval collection or numerical model-ing, and the resulting matrix can be hard to interpret.By focusing on the dynamics of upstream boundaries,our model helps to isolate the local aspects of connec-tivity that are most important to defining the spatialextent of a species on a regional scale. The summarystatistic used in Eq. (3) to discover the relative com-petiveness needed to maintain a boundary, the locallarval retention ƒret, can be used to analyze connec-tivity matrices or local larval settlement data tounderstand their implication for range boundaries.The patterns of species range boundaries we wouldpredict from estimates of larval retention ƒret in theGulf of Maine/Mid-Atlantic Bight are consistent withob served biogeographic transition zones for specieswith long planktonic durations.

For coastal benthic species whose dispersal is dom-inated by ocean currents, we find that the locations ofupstream boundaries are determined by the inter -action of larval retention and environmentally medi-ated gradients in relative competitiveness. The up -stream extent of a species range is most likely tooccur where either or both are maximized. No similarconstraint exists for the downstream extent of arange. Because the oceanographic features associ-ated with regions of enhanced retention (increasedƒret) are also associated with enhanced environmen-tal gradients (Savidge et al. 2013), we expect the up -stream extent of species ranges to cluster at theselocations. We expect that as climate changes, rangesof species with long-distance dispersing planktoniclarvae will remain fixed at one such location untilthey rapidly shift to the next; thus regions with anenhanced frequency of upstream species rangeboundaries should be monitored for signs of shiftingranges.

Acknowledgements. This work would not have been possi-ble without the efforts of James Manning and the NOAANortheast Fisheries Science Center in maintaining the Gulfof Maine drifter program; the data are available atwww.nefsc. noaa. gov/drifter/. Y. Li and K. Chen providedvaluable ocean circulation model data that guided our earlythinking about connectivity patterns. This work was fundedby NSF grants OCE 0961830, 0961344, and 1029841 andNOAA grant NA11NOS0120033. A persistent reviewer ledus to greatly increase the clarity of our presentation.

LITERATURE CITED

Aiken CM, Navarrete SA (2014) Coexistence of competi-tors in marine metacommunities: environmental varia -bility, edge effects, and the dispersal niche. Ecology 95:2289− 2302

Altman S, Robinson JD, Pringle JM, Byers JE, Wares JP(2013) Edges and overlaps in Northwest Atlantic phylo-geography. Diversity 5: 263−275

Aretxabaleta AL, McGillicuddy DJ Jr, Smith KW, Lynch DR(2008) Model simulations of the Bay of Fundy Gyre: 1. Climatological results. J Geophys Res 113: C10027

Aretxabaleta AL, McGillicuddy DJ, Smith KW, Manning JP,Lynch DR (2009) Model simulations of the Bay of FundyGyre: 2. Hindcasts for 2005−2007 reveal interannualvariability in retentiveness. J Geophys Res 114: C004948

Berkley HA, Kendall BE, Mitarai S, Siegel DA (2010) Turbu-lent dispersal promotes species coexistence. Ecol Lett 13: 360−371

Bode M, Bode L, Armsworth PR (2011) Different dispersalabilities allow reef fish to coexist. Proc Natl Acad SciUSA 108: 16317−16321

Byers JE, Pringle JM (2006) Going against the flow: reten-tion, range limits and invasions in advective environ-ments. Mar Ecol Prog Ser 313: 27−41

Case TJ, Taper ML (2000) Interspecific competition, envi-ronmental gradients, gene flow, and the coevolution ofspecies’ borders. Am Nat 155: 583−605

Case TJ, Holt RD, McPeek MA, Keitt TH (2005) The commu-nity context of species’ borders: ecological and evolu-tionary perspectives. Oikos 108: 28−46

Connolly SR, Roughgarden J (1999) Increased recruitmentof Northeast Pacific barnacles during the 1997 El Niño.Limnol Oceanogr 44: 466−469

Cowen RK, Paris CB, Srinivasan A (2006) Scaling of connec-tivity in marine populations. Science 311: 522−527

DiBacco C, Levin LA (2000) Development and application ofelemental fingerprinting to track the dispersal of marineinvertebrate larvae. Limnol Oceanogr 45: 871−880

Eckhart VM, Geber MA, Morris WF, Fabio ES, Tiffin P,Moeller DA, McPeek EMA (2011) The geography ofdemography: long-term demographic studies and spe-cies distribution models reveal a species border limitedby adaptation. Am Nat 178: S26−S43

Ewers-Saucedo C, Pringle JM, Sepúlveda HH, Byers JE,Navarrete SA, Wares JP (2016) The oceanic concordanceof phylogeography and biogeography: a case study inNotochthamalus. Ecol Evol 6: 4403−4420

Gaines SD, Lester SE, Eckert G, Kinlan BP, Sagarin R, Gay-lord B (2009) Dispersal and geographic ranges in the sea.In: Witman J, Roy K (eds) Marine macroecology. Univer-sity of Chicago Press, Chicago, IL, p 227–249

Gaylord B, Gaines SD (2000) Temperature or transport?

39

Mar Ecol Prog Ser 567: 29–40, 2017

Range limits in marine species mediated solely by flow.Am Nat 155: 769−789

Gotelli NJ (1991) Metapopulation models: the rescue effect,the propagule rain, and the core-satellite hypothesis. AmNat 138: 768−776

Hannah CG, Shore JA, Loder JW, Naimie CE (2001) Sea-sonal circulation on the Western and Central ScotianShelf. J Phys Oceanogr 31: 591−615

Hare MP, Guenther C, Fagan WF (2005) Nonrandom larvaldispersal can steepen marine clines. Evolution 59: 2509−2517

Harley CDG, Randall Hughes A, Hultgren KM, Miner BGand others (2006) The impacts of climate change incoastal marine systems. Ecol Lett 9: 228−241

Hutchins LW (1947) The bases for temperature zonation ingeographical distribution. Ecol Monogr 17: 325−335

Lentz SJ (2008) Observations of the mean circulationover the Middle Atlantic Bight continental shelf. J PhysOceanogr 38: 1203−1221

Lentz SJ (2010) The mean along-isobath heat and salt bal-ances over the Middle Atlantic Bight continental shelf.J Phys Oceanogr 40: 934−948

Limeburner R, Beardsley RC (1982) The seasonal hydro -graphy and circulation over Nantucket Shoals. J Mar Res40: 371−406

Locarnini RA, Mishonov AV, Antonov JI, Boyer TP and others(2013) World ocean atlas 2013, Vol 1: temperature. In: Levi-tus S, Mishonov A (eds) NOAA Atlas NESDIS 73. NOAA

Lutscher F, McCauley E, Lewis MA (2007) Spatial patternsand coexistence mechanisms in systems with unidirec-tional flow. Theor Popul Biol 71: 267−277

Mace AJ, Morgan SG (2006) Larval accumulation in the leeof a small headland: implications for the design of marinereserves. Mar Ecol Prog Ser 318: 19−29

Manning JP, McGillicuddy DJ Jr, Pettigrew N, Churchill JH,Incze LS (2009) Drifter observations of the Gulf of MaineCoastal Current. Cont Shelf Res 29: 835−845

Nagylaki T (1978) Clines with asymmetric migration. Genetics 88:813

Pappalardo P, Pringle JM, Wares JP, Byers JE (2015) Thelocation, strength, and mechanisms behind marine bio-geographic boundaries of the east coast of North Amer-ica. Ecography 38: 722−731

Pelc RA, Warner RR, Gaines SD (2009) Geographical patterns of genetic structure in marine species with con-trasting life histories. J Biogeogr 36: 1881−1890

Phillips SJ, Anderson RP, Schapire RE (2006) Maximumentropy modeling of species geographic distributions.Ecol Model 190: 231−259

Pinsky ML, Worm B, Fogarty MJ, Sarmiento JL, Levin SA(2013) Marine taxa track local climate velocities. Science341: 1239−1242

Pringle JM, Wares JP (2007) Going against the flow: mainte-nance of alongshore variation in allele frequency in acoastal ocean. Mar Ecol Prog Ser 335: 69−84

Pringle JM, Lutscher F, Glick E (2009) Going against theflow: effects of non-Gaussian dispersal kernels andreproduction over multiple generations. Mar Ecol ProgSer 377: 13−17

Pringle JM, Blakeslee AMH, Byers JE, Roman J (2011)Asymmetric dispersal allows an upstream region to con-trol population structure throughout a species’ range.Proc Natl Acad Sci USA 108: 15288−15293 6

Robinson A, Brink KH (2006) The sea, Vol 14A, B. The globalcoastal ocean: interdisciplinary regional studies and syn-thesis. Harvard University Press, Cambridge, MA

Salomon Y, Connolly SR, Bode L (2010) Effects of asymmet-ric dispersal on the coexistence of competing species.Ecol Lett 13: 432−441

Savidge DK, Austin JA, Blanton BO (2013) Variation in theHatteras Front density and velocity structure. Part 2: His-torical setting. Cont Shelf Res 54: 106−116

Sexton JP, McIntyre PJ, Angert AL, Rice KJ (2009) Evolutionand ecology of species range limits. Annu Rev Ecol EvolSyst 40: 415−436

Siegel DA, Kinlan BP, Gaylord B, Gaines SD (2003) La -grangian descriptions of marine larval dispersion. MarEcol Prog Ser 260: 83−96

Small ST, Wares JP (2010) Phylogeography and marineretention. J Biogeogr 37: 781−784

Spalding MD, Fox HE, Allen GR, Davidson N and others(2007) Marine ecoregions of the world: a bioregionaliza-tion of coastal and shelf areas. Bioscience 57: 573−583

Strathmann MF (1987) Reproduction and development ofmarine invertebrates of the northern Pacific Coast: dataand methods for the study of eggs, embryos, and larvae.University of Washington Press, Seattle, WA

Sunday JM, Pecl GT, Frusher S, Hobday AJ and others(2015) Species traits and climate velocity explain geo-graphic range shifts in an ocean-warming hotspot. EcolLett 18: 944−953

Vander Woude AJ, Largier JL, Kudela RM (2006) Nearshoreretention of upwelled waters north and south of PointReyes (northern California) — patterns of surface tem-perature and chlorophyll observed in CoOP WEST. DeepSea Res II 53: 2985−2998

Wares JP (2002) Community genetics in the NorthwesternAtlantic intertidal. Mol Ecol 11: 1131−1144

40

Editorial responsibility: Claire Paris, Miami, Florida, USA

Submitted: May 23, 2016; Accepted: January 23, 2017Proofs received from author(s): February 26, 2017