successional mechanism varies along a gradient in ...hydrothermal fluid flux at deep-sea vents...

523

Ecological Monographs, 73(4), 2003, pp. 523–542q 2003 by the Ecological Society of America

SUCCESSIONAL MECHANISM VARIES ALONG A GRADIENT INHYDROTHERMAL FLUID FLUX AT DEEP-SEA VENTS

LAUREN S. MULLINEAUX,1 CHARLES H. PETERSON,2 FIORENZA MICHELI,3 AND SUSAN W. MILLS1

1Biology Department, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts 02543 USA2University of North Carolina at Chapel Hill, Institute of Marine Sciences, Morehead City, North Carolina 28557 USA

3Hopkins Marine Station, Stanford University, Pacific Grove, California 93950 USA

Abstract. Invertebrate communities inhabiting deep-sea hydrothermal vents undergosubstantial succession on time scales of months. Manipulative field experiments assessedthe relative roles of environmental state and biotic interactions in determining temporalsuccession along a spatial gradient in vent fluid flux at three vent sites near 98509 N on theEast Pacific Rise (2500 m water depth). Species colonization patterns on cubic basalt blocks(10 cm on a side) deployed by the submersible Alvin revealed both positive (facilitation)and negative (inhibition) biological interactions, in the context of established successiontheory. Over a series of four cruises from 1994 to 1998, blocks were exposed to colonistsfor consecutive and continuous intervals in short-term (5 1 8 5 13 mo) and longer-term(8 1 29 5 37 mo) experiments. Colonists grouped into a mobile functional group wereless abundant in the continuous interval (13 mo) than in the synchronous pooled-consecutiveintervals (5 1 8 mo) of the short-term experiment, indicating that early colonists inhibitedsubsequent recruitment. Colonists grouped into a sessile functional group exhibited theopposite pattern, indicating facilitation. Similar trends, though not statistically significant,were observed in the longer-term experiment. The character of species interactions variedalong a gradient in hydrothermal fluid flux (and inferred productivity), with inhibitoryinteractions more prominent in zones with high temperatures, productivity, and faunaldensities, and facilitative interactions appearing where temperatures, productivity, and den-sities were low. Analyses of primary succession on introduced basalt blocks suggest thatbiological interactions during early vent community development strongly modify initialpatterns of settlement, even in the absence of sustained temporal change in the vent fluidflux.

Key words: deep sea; facilitation; hydrothermal vent; inhibition; Lepetodrilus; limpet; Ophry-otrocha; polychaete; productivity gradient; succession; tubeworm; vestimentiferan.

INTRODUCTION

The pathways of ecological succession, defined as adirectional change in community structure following aperturbation that creates new or modifies existing hab-itat, may be fundamental characteristics of most ani-mal, plant, and microbial communities. Uncovering themechanisms driving these temporal sequences of spe-cies replacements and community composition shiftsis essential for a full understanding of the structure,diversity, and dynamics of biotic systems. Although afew frequently cited early studies suggested that suc-cession in some systems might progress unidirection-ally toward an inevitable and fixed climax (e.g., Cle-ments 1916; reviewed in Odum 1969), most subsequentempirical and theoretical investigations indicate thatthe trajectories vary and can be difficult to predict. Forinstance, during primary succession in benthic marinesystems, very different outcomes can occur, dependingon the identity of initial colonists (Sutherland and Karl-son 1977), the relative importance of positive and neg-

Manuscript received 28 October 2002; accepted 26 November2002; final version received 19 December 2002. CorrespondingEditor: R. J. Etter.

ative interactions among species (Sousa 1979a, b, Lub-chenco 1983), and the role of indirect effects of con-sumers (Hixon and Brostoff 1996). Further complica-tions occur when the abiotic environment variesunpredictably, or during secondary succession after adisturbance that has not completely eradicated the ex-isting community (i.e., leaving a biological legacy thatinfluences the course of species reestablishment, as oc-curred at Mt. St. Helens after the 1980 eruption; Frank-lin and MacMahon 2000).

Succession is particularly apparent and important insystems where episodic disturbances induce largechanges in availability of primary substratum or food.Such dramatic disturbances occur at deep-sea hydro-thermal vents, where catastrophic events such as vol-canic eruptions and tectonic events (quakes) create ordestroy entire habitats. On fast-spreading mid-oceanridges, catastrophic perturbations occur on time scalesof years to decades (MacDonald et al. 1980, Haymonet al. 1993), making disturbance and primary succes-sion dominant characteristics of the ecosystem. Thephysical and chemical nature of the habitat continuesto be dynamic after eruptions, with rerouting of hy-drothermal fluid sources due to tectonic shifts or geo-

524 LAUREN S. MULLINEAUX ET AL. Ecological MonographsVol. 73, No. 4

chemical clogging, changes in chemical compositionof fluids (von Damm 1995, Butterfield et al. 1997) androck falls or basaltic pillow collapses (Haymon et al.1993). These physical perturbations modify the habitat,and the chemical variations (i.e., changes in hydro-thermal fluid flows that support the chemoautotrophic-based food web) alter supplies of the primary nutri-tional source. The responses of colonizing species atdeep-sea vents to these perturbations, either directlythrough physiological tolerances and nutritional re-quirements, or indirectly through biological interac-tions, are largely unknown.

The mechanisms of succession in the vent environ-ment may parallel those in forests, streams, or shallowmarine environments, or they may differ greatly dueto the unique combination of habitat transience andpatchiness, species endemism, and steep gradients intemperature, toxic chemicals, and chemoautotrophicproduction. Given the insights that successional studieshave provided into the ecology of intertidal (Dayton1971, Sousa 1979a, b), subtidal (Witman 1987), anddeep-sea (Grassle and Morse-Porteous 1987) commu-nities, understanding successional change in hydro-thermal vent communities is necessary to explain theirspatial and temporal patterns and may simultaneouslyprovide broader insights into the general nature of eco-logical succession.

Previous studies of invertebrate colonization at hy-drothermal vents have documented a distinct and con-sistent successional sequence over the lifespan of ventson the East Pacific Rise (Fustec et al. 1987, Hessler etal. 1988). The initial visibly dominant sessile metazoanin sites with moderate-temperature fluids (i.e., ,308Canomaly; not high-temperature black smokers) appearsto be the small vestimentiferan tubeworm Tevnia jer-ichonana (Shank et al. 1998). This species then is re-placed by the larger tubeworm Riftia pachyptila, fre-quently over a period of less than one year. Later inthe sequence, the mussel Bathymodiolus thermophiluscolonizes and may, in some cases, displace the vesti-mentiferans (Hessler et al. 1988). Time-series obser-vations suggest that this sequence can correspond to achange in the temperature and chemistry of vent fluids(Shank et al. 1998), although Hessler et al. (1988) andMullineaux et al. (2000) document species successionunder conditions of constant venting and conclude thatbiological interactions also play a role. Because moststudies of vent ecology are observational and tend totarget the larger species, the mechanisms causing ob-served successional replacements remain speculative,and patterns for a vast number of smaller, but numer-ically important, species (i.e., gastropods, polychaetes)are unexplored. The alternative approach is controlledmanipulative experimentation, which has proven usefulin elucidating mechanisms of succession in other ma-rine environments (e.g., Dayton 1971, Sutherland andKarlson 1977, Sousa 1979a, b, Farrell 1991, Benedetti-Cecchi 2000).

The objective of our study was to characterize andquantify changes over time in the invertebrate com-munities colonizing new surfaces at hydrothermalvents along a gradient in vent fluid flux, and to examinewhat mechanisms may be effecting these changes. In-corporating successional theory from marine and ter-restrial systems (starting with Connell and Slatyer1977; modified with subsequent studies as reviewed inMcCook 1994), and documenting colonization of newhabitat during staggered, but overlapping, observationintervals (as in Ambrose 1984), we reasoned that faunalchanges at vents are driven by one or more of thefollowing processes:

1) Inhibition. The initial invertebrate colonists in-teract with later arrivals in a way that deters subsequentcolonization. Under these conditions, we expect thatcolonists of a new habitat exposed over an extendedinterval would be fewer than the sum of colonists onsimilar habitats exposed sequentially over the sametime period. The extreme case for inhibition is pre-emption, in which the initial colonists prevent futuresettlers or immigrants from becoming established (asdescribed in Underwood and Denley 1984), and main-tain their dominance for an extended period. In thiscase, the numbers and species composition of colonistson habitats exposed at the same time would be similar,regardless of the duration of exposure (presuming theabsence of any external perturbation to restart succes-sion).

2) Facilitation. The initial colonists interact with lat-er arrivals in a way that enhances subsequent coloni-zation. In this case, we expect that colonists on a habitatexposed over an extended interval would be more nu-merous and more diverse than the sum of colonists onsimilar habitats exposed sequentially over the same in-terval. The extreme case for facilitation is obligate fa-cilitation (Odum 1969), in which a secondary speciescould colonize only after a pioneer species became es-tablished.

3) Tolerance. Initial colonists interact weakly or notat all with subsequent arrivals, so colonists accumulateas propagules arrive independently of the species com-position or abundance of pioneers (Egler 1954). In thiscase, we expect that colonists on a habitat exposed overan extended interval would be equivalent to the sumof colonists on similar habitats exposed sequentiallyover the same interval. This scenario serves as our nullmodel.

In addition to the above inhibition and facilitationmodels, which assume that species interactions drivethe successional sequence toward a specific trajectory(canalized succession; Berlow 1997), the sequence ofspecies replacement might vary in space and time de-pending on the variation in the physical characteristicsof the environment and in the composition of the avail-able larval pool (externally driven succession; Berlow1997). In the highly dynamic vent habitat, successionalchanges may reflect species’ responses to variation in

November 2003 525SUCCESSION AT DEEP-SEA HYDROTHERMAL VENTS

FIG. 1. Map of East Pacific Rise axis (shaded corridor)near 98509 N, showing Biovent, East Wall, and Worm Holevent sites between the Clipperton (C) and Sequieros (S) frac-ture zones.

the environment (i.e., any of the thermal, geochemical,or nutritional properties that covary with vent fluidflux), as well as species’ interactions with each other.Because it is important to distinguish between thesemechanisms (e.g., Callaway and Walker 1997), wemonitored both hydrothermal vent fluid flux (usingtemperature as a proxy; Johnson et al. 1988, 1994) andthe ambient faunal community throughout the study.Also, because environmental and faunal variation oc-curs among, as well as within, individual vent fields,we conducted our studies at several different sites withdistinctly different community structure, and over agradient of vent fluid flux at each of these sites. Inaddition, we carried out the experiments over both short(months) and longer (years) time scales in order toevaluate interactions that may occur on different tem-poral scales. This temporal perspective becomes im-portant when different mechanisms operate at differentstages of succession (as reviewed by Caswell 1976).

It is important to recognize that change in speciescomposition over the course of colonization of newhabitat may be mediated by mobile predators that arenot typically monitored as part of a successional study.For instance, an increase in abundance of a colonizingspecies may result from interactions with a facilitatingspecies, or indirectly by reduction of a competing spe-cies through predation (as in Hixon and Brostoff 1996).Although the activities of large mobile predators werenot monitored directly in the present study, informationon their activities and diets is available from associatedprojects in vent habitats (Micheli et al. 2002; C. H.Peterson, unpublished data; G. Sancho, unpublisheddata).

MATERIALS AND METHODS

Study sites

Studies were conducted at vents near 98509 N,1048179 W along the East Pacific Rise (EPR; Fig. 1)at a water depth of 2500 m. This location was selectedbecause nearby vent communities had been monitoredfrequently since 1991, when a well-documented vol-canic eruption occurred (Haymon et al. 1993). Thethree vent fields selected appeared to be at differentstages in the colonization sequence (as defined byShank et al. 1998), with Worm Hole at an early stage(vestimentiferan clumps dominated by T. jerichonana),East Wall at a midstage (vestimentiferan clumps co-occupied by T. jerichonana and R. pachyptila; see Plate1), and Biovent at a later stage (vestimentiferan clumpsinhabited by R. pachyptila with co-occurring mussels).

At each site, the community inhabited an area thatspanned a spatial gradient of hydrothermal fluid flux.With the exception of organisms inhabiting one focused‘‘black smoker’’ orifice at Biovent (where hydrother-mal fluids reached temperatures of 4038C, and con-tained concentrations of H2S of up to 30 mmol/L; vonDamm 2000), our three study sites experienced vent

fluids that had been released from the seafloor aftermixing with ambient seawater (28C) and exhibited tem-perature anomalies in the range of 308C to ,18C. Be-cause the high-temperature fluids were characterizedby low pH and potentially toxic concentrations of bothH2S and heavy metals, a gradient in fluid flux representsa gradient in potential thermal and chemical stresses.The fluid flux gradient also is assumed to represent aproductivity gradient because chemosynthetic produc-tion depends on the reduced chemicals in the vent flu-ids. This assumption is supported by the positive re-lation observed between vent fluid flux and faunal den-sity and cover (C. Fisher, unpublished data).

The faunas at these sites displayed a striking bio-logical zonation along the vent fluid flux gradient thatis characteristic of vents in this region of the EPR (Hes-sler et al. 1985): alvinellid polychaetes and associatedspecies inhabit areas of high-temperature vent flow(.258C anomaly); vestimentiferan tubeworms occur invigorous flow (up to 25–308C above ambient); bivalves(mussels and clams) inhabit regions with moderate flow(up to 108C above ambient); and various suspensionfeeders, including barnacles and serpulid polychaetes,inhabit weak flow (,28C above ambient). The Bioventsite exhibited all these zones, whereas the East Walland Worm Hole sites did not exhibit black smokers (orthe associated alvinellid residents) and at Worm Hole,the suspension-feeders were intermixed with sparse,juvenile mussels. For our colonization study, we ex-cluded the black smoker habitat and designated theother distinct faunal zones by the prominent signaturespecies as vestimentiferan, bivalve, and suspensionfeeder. At Worm Hole, the region with mixed bivalvesand suspension feeders experienced moderate ventflows, but lacked the dense beds of mature mussels, sowas designated as an intermediate bivalve/suspension-


PLATE 1. Recovery of a basalt colonization block by the submersible Alvin from an East Pacific Rise vent communityin a mid-successional stage (i.e., vestimentiferan zone dominated by the tubeworm Riftia pachyptila with no overgrowth bymussels). (Left) Limpets and polychaetes were included in analyses of colonization; large transient organisms such as thecrab Bythograea thermydron were not. Limpets are also visible on the surface of vestimentiferan tubes in the background.(Right) The manipulator claw of the submersible Alvin is used to conduct studies of ecological succession at deep-seahydrothermal vents. Clumps of the tubeworm Riftia pachyptila are typical inhabitants of mid-successional communitiesinhabiting zones of moderate-temperature venting (58–258C). Markers identifying experiments are visible in the background.

feeder zone. We also defined at all sites a peripheryzone that lacked a detectable temperature anomaly andgenerally was occupied by deep-sea invertebrates withno known affinity to vents, except the vent-endemicforaminifer species Abyssotherma pacifica.

Experimental design

We tested for positive (facilitative) and negative (in-hibitory) species interactions by quantifying coloni-zation of vent species on cubic basalt blocks roughly10 cm on a side. The submersible Alvin was used toposition and recover the blocks in the appropriate hab-itats on the seafloor. For each of five deployment in-tervals, replicate blocks (n 5 3) were introduced intofour zones (vestimentiferan, bivalve, suspension feed-er, periphery) at each of three vent fields (Biovent, EastWall, and Worm Hole; at Worm Hole, the bivalve andsuspension-feeder zones were intermixed) locatedalong a 2-km section of the axial valley of the ridge.‘‘Interval’’ treatments were clustered in three replicateplots within each zone (Fig. 2). We were aware thatthe number of replicates was low, and might compro-mise our ability to detect treatment effects. However,given our lack of a priori knowledge of the naturalhistory of the vent ecosystem, we chose to spread ourstudies over multiple sites and zones, with the intentof discovering trends that occurred generally over thetemporal and spatial mosaic of vents on the EPR.

The fluid microhabitat of each block was character-ized on deployment and recovery by measuring water

temperature with the Alvin temperature probe at thelowermost extremity of the block (as in Mullineaux etal. 2000). Longer-term (months to years) temperaturemeasurements were obtained with an internally re-cording ‘‘Hobo’’ temperature probe (Onset, Pocasset,Massachusetts, USA) at each zone for the duration ofthe block deployments. Ambient temperature of thebottom water was measured as 1.88C by the Alvinprobe, and as 2.08C by the lower-resolution ‘‘Hobo’’probe.

Sets of blocks were deployed and recovered on fourcruises (in November 1994, April 1995, December1995, May 1998) in a series of overlapping time in-tervals (November 1994 to April 1995, 5 mo; April1995 to December 1995, 8 mo; November 1994 to De-cember 1995, 13 mo; April 1995 to May 1998, 37 mo;December 1995 to May 1998, 29 mo). Blocks from thelongest possible interval (November 1994 to May1998, 42 mo) are not included in analyses because re-covery was poor. The purpose of these overlappingdeployments was to test for positive or negative bio-logical interactions, in the context of succession theory,as modified from Connell and Slatyer (1977). The ra-tionale followed that of Ambrose (1984): if the sum ofthe colonists on blocks in two consecutive (i.e., 5 and8 mo) intervals is the same as the number in the con-tinuous (13 mo) interval, then initial colonists wouldappear to have no detectable effect on later arrivals,supporting our null model (tolerance) of no effect ofbiological interactions on the successional sequence. If


FIG. 2. Diagram (not to scale) of experimental design at a single vent site showing plots of colonization blocks withvarious ‘‘interval’’ treatments (5, 8, 13, 29, and 37 mo) within each zone.

the sum of colonists across the consecutive intervalsis less than the numbers accumulated over the contin-uous interval then positive, facilitative interactionswould be indicated. If, however, the sum of the colo-nists in the consecutive intervals is greater, inhibitoryinteractions, such as space preemption or competitionfor nutritional resources, would be suggested. Separateanalyses were conducted for shorter (13 mo) and longer(37 mo) study intervals.

Blocks were recovered by the submersible Alvin andplaced in individual collection compartments for trans-port back to the ship, as in Mullineaux et al. (2000).On shipboard, blocks and their attached colonists werepreserved in 80% ethanol, as were any detached in-dividuals from the compartment retained on a 63-mmsieve. In the laboratory, each block was examined un-der a dissecting microscope; all metazoan and selectedprotozoan (foraminifer and folliculinid ciliate) colo-nists were enumerated and identified (to species whenpossible). Folliculinid abundances were estimated fromsubsamples when densities exceeded 100 individualsper block face. From the sieved samples, all individuals.1 mm in length were counted and identified. Thesmaller individuals, mostly juvenile gastropods and

polychaetes, were very numerous but too difficult toidentify to species using standard morphological tech-niques.

Species were grouped into mobile and sessile func-tional groups, based on their morphologies and onknown behaviors of related species (Table 1). Poolingindividual species into functional categories decreasedvariability among replicate blocks and increased thepower of statistical tests. Analysis of functional groupsin the short-term (13 mo) experiment was performedwith a three-way nested MANOVA (Systat version 10,Systat, Evanston, Illinois, USA), testing effects of ‘‘in-terval,’’ ‘‘zone,’’ and ‘‘site’’ on sessile and mobile spe-cies groups. Intervals were categorized into two levels:pooled-consecutive, with colonists summed from the5-mo and 8-mo blocks within each plot, and continuous(13 mo). A split-plot analysis (with intervals clusteredinto plots within zones) was originally planned for thisdesign, but a nested approach was used instead becauseof the complexity of the multifactor MANOVA. Pre-liminary tests comparing the nested and split-plot ap-proaches within individual sites showed no differencein detection of significant interval and zone effects.The zone factor had three levels (vestimentiferan, sus-


TABLE 1. Mean numbers of individuals in taxa colonizing blocks over the short-term exposures (5-, 8- and 13-mo) atBiovent, East Wall, and Worm Hole, and long-term exposures (8-, 29- and 37-mo) at East Wall.

Taxon Taxonomic group Fn Grp

Biovent13 mo

(n 5 33)

East Wall13 mo

(n 5 30)

Worm Hole13 mo

(n 5 27)

East wall37 mo

(n 5 33) ANOVA

Metafolliculina sp.Lepetodrilus elevatusAbyssotherma pacificaAmphisamytha galapagensisVestimentiferansLaminatubus alviniLepetodrilus pustulosusOphryotrocha akessoni

ciliategastropodforaminiferanpolychaetevestimentiferanpolychaetegastropodpolychaete

SMSSSSMM

54.4565.2416.7923.1516.55

2.738.762.24

451.5724.2726.0016.8024.87

2.738.404.83

818.31186.39

51.6725.6438.6119.9719.08

4.42

834.2122.27

146.6168.6120.0015.27

3.248.18

prominentprominentsignature†prominentsignature†signature†prominentprominent

Cyathermia natacoidesVentiella sulfurisBathymodiolus thermophilus

gastropodamphipodbivalve

MMS

8.975.002.03

12.1717.87

1.37

4.894.583.42

3.302.218.61 signature

Lepetodrilus ovalisClypeosectus delectusArchinome rosaceaEulepetopsis vitreaParalvinella spp?Gorgoleptis spiralisLepetodrilus cristatusNereis sp.

gastropodgastropodpolychaetegastropodpolychaetegastropodgastropodpolychaete

MMMMSMMM

1.391.331.330.362.060.151.030.15

0.572.231.230.304.070.101.200.47

9.335.081.615.330.002.250.000.92

8.553.825.943.240.120.520.520.85

························

Polynoid polychaetesBathymargarites symplectorHesiospina vestimentiferaNeolepas zevinaeNicomache sp.Unsegmented wormGalapagomystides aristataOphiuroids

polychaetegastropodpolychaetebarnaclepolychaetemainly platyhelminthspolychaeteophiuroid

MMMSSMMM

0.270.360.090.180.060.120.820.12

0.470.330.000.170.000.100.000.27

0.530.140.000.420.190.810.000.33

0.791.001.420.701.030.180.060.12

························

Helicoradomenia acredemaLepidonotopodium spp.Tanaids

aplacophoranpolychaetecrustacean

MMS

0.060.000.09

0.030.130.40

0.140.500.08

0.550.090.12

·········

Bathypecten vulcaniIphionella risensisCnidarianBrown papillated wormBranchipolynoe sp.Gorgoleptis emarginatusBarnacle cypridsGlycera sp.

bivalvepolychaetecnidarianpolychaetepolychaetegastropodbarnaclepolychaete

SMSMMMSM

0.120.150.000.030.090.000.000.03

0.230.130.470.000.000.170.300.07

0.060.000.000.140.030.000.030.06

0.150.210.000.180.240.180.000.09

························

Dorvilleid polychaeteAnemoneRhynchopelta concentricaProvanna sp.Prionospio sandersiIsopodsPeltospira delicataDepressigyra planispira

polychaeteanemonegastropodgastropodpolychaetecrustaceangastropodgastropod

MSMMMMMM

0.030.000.000.000.000.030.000.00

0.030.030.000.000.070.070.030.00

0.110.080.000.000.000.000.000.00

0.030.090.180.120.060.000.000.03

························

Hesionid polychaeteSmall gastropodsSmall limpets

polychaete MMM

0.003.481.64

0.0317.8710.07

0.0033.03

3.36

0.000.03

36.52

·········

Gastropod juvenilesUnidentified gastropodPolychaete juvenilesUnidentified polychaetesUnidentified foraminifersLeptostracansTotal mobileTotal sessileTotal individuals

gastropodgastropodpolychaetepolychaeteforaminiferancrustacean

MMMMSMMS

5.120.000.060.090.060.00

98.45118.27216.73

27.930.000.600.030.730.00

86.27529.83616.10

36.390.030.530.080.280.19

107.671096.031203.70

36.554.790.240.120.520.00

279.11958.75

1237.86

··················

Fn GroupFn Group

Notes: Taxa are categorized into functional mobility groups (Fn Grp), based on inferences of predominant lifestyle madefrom morphology: M, mobile; S, sessile. Prominent colonizing taxa (among top 10 in at least three of the four site/exposurecombinations) and signature species for each zone (vestimentiferans, Bathymodiolus thermophilus, Laminatubus alvini, andAbyssotherma pacifica) were analyzed individually by ANOVA.

† Signature species also prominent.


pension feeder, and periphery), with blocks of differentintervals grouped into three plots within each zone.This analysis treated the bivalve/suspension-feederzone at Worm Hole as comparable to a suspension-feeder zone, because it lacked the distinct three-di-mensional habitat structure provided by the deep bi-valve cover at the other two sites. The bivalve zonesof Biovent and East Wall were excluded from the anal-ysis because no comparable zone occurred at WormHole. A two-way nested MANOVA was used for thelonger-term (37 mo) study at East Wall, with interval(two levels) and zone (four levels) as factors. Univar-iate ANOVAs were used to test the effect of factorsthat were significant in the MANOVAs. To determinewhich species contributed to the patterns exhibited bythe functional groups, similar univariate ANOVAswere used to test treatment effects on abundant indi-vidual species. Prior to statistical analysis, the datawere plotted and variances tested for homogeneity (us-ing Cochran’s tests at a 5 0.05) and for normal dis-tribution; logarithmic transformations, ln(x 1 1), wereperformed if necessary. Treatment means were com-pared using Tukey’s post hoc comparisons. Sign tests(Sokal and Rohlf 1995) were used to evaluate Intervaleffects in selected zones.

RESULTS

Colonization in continuous vs.pooled-consecutive intervals

Of the 112 blocks recovered and analyzed, all butfive had been colonized by vent-associated inverte-brates, with metazoan densities (excluding swimmingcrustaceans) ranging up to 2271 individuals per block,and total densities (including foraminifers and ciliates)reaching .7000. On most blocks, less than half theavailable surface area was colonized, although a fewwere very heavily colonized by vestimentiferans ormussel byssal threads and had .90% of the availablesurface covered. Individuals were separated into 55species or species-groups (Table 1) and assigned tofunctional mobility groups based on field observations,functional morphology or knowledge of related spe-cies. Colonists of the mussel B. thermophilus were as-signed to the sessile group because they are attachedwith byssal threads, and did not function as mobilegrazers like the other taxa assigned to the mobile group.

In the short-term experiment, the mobile and sessilegroups showed very different patterns in response tointerval and zone treatments. MANOVA revealed thatthe interval, zone, and site effects on the functionalgroups were significant (P , 0.05; Table 2). ANOVAshowed a significant interval effect for the mobilegroup, which was more abundant in the pooled-con-secutive than the continuous intervals (P 5 0.013; Ta-ble 2). In contrast, the sessile group tended to be moreabundant in the continuous than in pooled-consecutiveintervals (Fig. 3), but this result was not significant

(Table 2). The interval effect on each functional groupwas most consistent in the zone where that group wasmost abundant; i.e., the mobile group in the vestimen-tiferan zone and the sessile group in the suspension-feeder zone (Fig. 3). ANOVA also showed significantzone and site effects and a zone 3 site interaction inboth mobile and sessile groups. Post hoc Tukey testsshowed that abundances of both groups generally de-creased with decreasing fluid flux, with significantlyhigher abundances in the vestimentiferan zone than inthe periphery (Table 2). Abundances in the vestimen-tiferan zone were significantly higher than in the sus-pension-feeder zone at Biovent and East Wall, but notWorm Hole.

In the longer-term experiment, MANOVA detecteda significant effect of zone (P 5 0.02; Table 2), butnot interval, on the functional groups. ANOVA showedthat this zone effect was significant for both mobileand sessile groups, with abundances in the peripheryzone lower than were those in the other zones. ANOVAalso showed that sessile invertebrate abundances dif-fered significantly among the plots.

Nine taxa were selected (Table 1) to evaluate theirindividual contribution to the interval effects. Taxawere selected for analysis if they were among the 10most abundant colonizing species in at least three ofthe four site/exposure data sets of Table 1 (the limpetsLepetodrilus elevatus and Lepetodrilus pustulosus, thepolychaetes Ophryotrocha akessoni, Laminatubus al-vini, and Amphisamytha galapagensis, vestimentifer-ans, the foraminifer Abyssotherma pacifica, and the cil-iate Metafolliculina sp.), or if they were a signatureadult resident of a zone (the mussel Bathymodiolusthermophilus). The ‘‘vestimentiferans’’ were treated asa group because small (,1 mm) individuals of the threespecies, Riftia pachyptila, Tevnia jerichonana, andOasisia alvinae, could not be identified reliably to spe-cies using morphological techniques. Species levelidentification of small individuals is possible using mo-lecular genetic techniques (as in Mullineaux et al.2000), but that approach was not feasible for the presentstudy because of the large number of vestimentiferancolonists.

In both the short- and longer-term experiments, afew individual species showed interval effects thatwere very similar to their corresponding functionalgroup (Figs. 4–6). In the short-term study, three mobileinvertebrate species, the limpets L. elevatus and L. pus-tulosus and the polychaete O. akessoni (except at Bio-vent), tended to be more abundant in the pooled-con-secutive (5 1 8 mo) than the continuous (13 mo) in-tervals (Fig. 4). Two sessile invertebrate species, theprotozoans A. pacifica and Metafolliculina sp., dem-onstrated the opposite pattern, usually being moreabundant in the continuous (13 mo) than the pooled-consecutive (5 1 8 mo) interval. The polychaetes A.galapagensis and L. alvini, the vestimentiferans, andthe mussel B. thermophilus did not show interval pat-


TABLE 2. MANOVAs comparing abundance of individuals in mobile and sessile groups (A) between continuous (13-mo)and pooled-consecutive (5 1 8-mo) intervals, among three zones (vestimentiferan, suspension-feeder, and periphery), andamong three sites (Biovent, East Wall, and Worm Hole) in the short-term experiment; and (B) between continuous (37-mo) and pooled-consecutive (8 1 29-mo) intervals, among four zones (including bivalve) at East Wall in the longer-termexperiment.

Source df

Univariate ANOVA,mobile

MS F P

Univariate ANOVA,sessile

MS F P

MANOVA

Wilks’ l F P

A) Short-term (13 mo)IntervalZoneSiteInterval 3 ZoneZone 3 SiteInterval 3 SiteInterval 3 Zone 3 SitePlot (Zone)ErrorTotal sum of squares

12224246

2649

5.2390.5611.84

0.503.420.750.711.190.78

256.46

6.71116.18

15.190.654.390.960.911.53

0.02,0.01,0.01

0.530.010.400.480.21

0.2279.1841.47

1.834.031.230.720.781.34

306.17

0.1659.0630.94

1.363.000.920.540.59

0.69,0.01,0.01

0.270.040.410.710.74

0.73,0.01

0.250.900.480.850.81

5.0866.2313.37

0.732.991.170.77

0.01,0.01,0.01

0.580.010.370.63

B) Long-term (37 mo)IntervalZoneInterval 3 ZonePlot (Zone)ErrorTotal sum of squaresTukey test

13386

v s b p

2.029.390.601.091.35

48.80

1.506.980.450.81

0.270.020.730.62

,0.018.010.072.280.13

43.24s v b p

0.0363.21

0.5517.96

0.87,0.01

0.67,0.01

0.800.140.70

0.633.900.33

0.570.020.91

Notes: N 5 3 for most cases; exceptions are displayed in Fig. 3. Logarithmic transformations ln(x 1 1) were performedon data when necessary to obtain homogeneity of variances. For tests with a significant zone, site, or interval effect (P ,0.05), post hoc Tukey tests show means that are not significantly different (P . 0.05) connected with underlines (sites aredesignated as B, E, and W; zones as v, b, s, and p; intervals as C and P).

Tukey tests:Short-term mobile WvP EvP BvC BvP WvC WsP WsC EvC EsP BsP EsC EpP WpP BsC WpC BpP EpC BpCShort-term sessile Ws Wv Ev Bv Bs Es Wp Ep BpLong-term mobile v s b pLong-term sessile s v b p

terns that were consistent among sites. ANOVA indi-cated that the interval effect was significant for O. akes-soni (P , 0.05 without Bonferroni correction; Table3), as well as site and zone effects, and interval 3 zoneand zone 3 site 3 interval interactions. The intervalinteractions occurred because abundances were sig-nificantly higher in pooled-consecutive than contin-uous intervals in the suspension-feeder zone at allsites and in the vestimentiferan zone at East Wall andBiovent, but the opposite pattern occurred in the ves-timentiferan zone at Biovent, and the effect was notsignificant in the periphery (Tukey tests, P , 0.05;Table 3). Interval effects were not significant for anyof the other taxa.

Results from the longer-term experiment at East Wall(Figs. 4–6) were qualitatively similar to those of theshort-term studies. Three mobile species (L. elevatus,L. pustulosus, and O. akessoni) and the vestimentifer-ans tended to be more abundant in the pooled-consec-utive (8 1 29 mo) than in the continuous (37 mo)intervals. The sessile protozoan A. pacifica showed theopposite pattern in the suspension-feeder zone, whereit tended to be more abundant in the continuous inter-val, whereas Metafolliculina sp. did not differ detect-ably between the intervals. The three other prominent

taxa (L. alvini, A. galapagensis, and B. thermophilus)tended to be more abundant in the pooled-consecutiveintervals in the Vestimentiferan zone, and more abun-dant in the continuous interval in the suspension-feederzone (Figs. 4–6).

The power of our analyses of variance was low dueto limited replication and high variation among repli-cates. A sign test was used as an alternative method toevaluate Interval trends that were consistent acrosssites but not significant in the ANOVAs. In the sus-pension-feeding zone, mean sessile-group abundanceswere higher in the continuous than the pooled-consec-utive intervals at all three sites and during both ex-perimental durations. The probability of this patternoccurring in the absence of an interval effect would beP 5 0.0625. The same pattern occurred for the sessilespecies Metafolliculina sp., whereas the opposite pat-tern (abundances higher in pooled-consecutive thancontinuous intervals in all four cases) occurred in thevestimentiferan zone for the mobile species L. elevatusand L. pustulosus. The low probabilities of these pat-terns occurring in the absence of an Interval effectsuggest that our ANOVAs may have failed to detect areal effect for these groups.


FIG. 3. Abundance of mobile and sessile species-groups on blocks from pooled-consecutive (short-term, 5 1 8 mo; longer-term, 8 1 29 mo) and continuous (short-term, 13 mo; longer-term, 37 mo) intervals, at four zones (vestimentiferan, bivalve,suspension-feeder, and periphery) in three sites (Biovent, East Wall, and Worm Hole). At Worm Hole, two zones were mergedinto a bivalve/suspension-feeder zone. Error bars represent 1 SE for n 5 3, except as noted on figure (nd, no data). Resultsfrom a three-way MANOVA and post hoc Tukey tests are in Table 2.


FIG. 4. Abundance of gastropods Lepetodrilus elevatus and Lepetodrilus pustulosus and the polychaete Ophryotrochaakessoni on blocks from pooled-consecutive (short-term, 5 1 8 mo; longer-term, 8 1 29 mo) and continuous (short-term,13 mo; longer-term, 37 mo) intervals, at four zones (vestimentiferan, bivalve, suspension-feeder, and periphery) in three sites(Biovent, East Wall, and Worm Hole). Error bars represent 1 SE for n 5 3, except as noted in Fig. 3. Results from a three-way MANOVA and post hoc Tukey tests are in Table 3.

Variation of colonization among zones and sites

Colonists of two of the zonal signature species, ves-timentiferans and the serpulid polychaete L. alvini, set-tled in spatial patterns similar to the adults. Vestimen-

tiferan colonists were found almost exclusively in thevestimentiferan zone; the exceptions occurred on fourblocks in the bivalve or bivalve/suspension-feederzones (Fig. 4). Video images of the exceptional blocks


FIG. 5. Abundance of the polychaete Amphisamytha galapagensis and the mussel Bathymodiolus thermophilus on blocksfrom pooled-consecutive (short-term, 5 1 8 mo; longer-term, 8 1 29 mo) and continuous (short-term, 13 mo; longer-term,37 mo) intervals, at four zones (vestimentiferan, bivalve, suspension-feeder, and periphery) in three sites (Biovent, East Wall,and Worm Hole). Error bars represent 1 SE for n 5 3, except as noted in Fig. 3. Results from a three-way MANOVA andpost hoc Tukey tests are in Table 3.

at East Wall revealed previously undetected residentvestimentiferans within 1 m of each block, suggestingthat these blocks may have been placed in an anoma-lous microhabitat within that zone. The exceptional

block at Worm Hole had a measured temperature of5.08C, which was substantially higher than tempera-tures adjacent to the other blocks in the bivalve/sus-pension-feeder zone (1.8–3.88C), and within the range


FIG. 6. Abundance of the polychaete Laminatubus alvini, the foraminifer Abyssotherma pacifica, and the ciliate Meta-folliculina sp. on blocks from pooled-consecutive (short-term, 5 1 8 mo; longer-term, 8 1 29 mo) and continuous (short-term, 13 mo; longer-term, 37 mo) intervals, at four zones (vestimentiferan, bivalve, suspension-feeder, and periphery) inthree sites (Biovent, East Wall, and Worm Hole). Error bars represent 1 SE for n 5 3, except as noted in Fig. 3. Results froma three-way MANOVA and post hoc Tukey tests are in Table 3.

measured in the vestimentiferan zone at that site (1.9–10.98C). Colonists of the signature species for the sus-pension-feeder zone, the serpulid polychaete L. alvini,were more abundant in that zone than elsewhere.

In contrast, the signature species of the bivalve zone,B. thermophilus, did not colonize in highest numbersthere (Fig. 5). This anomalous pattern was most distinctat East Wall, where mussel adults were large and abun-


dant in the bivalve zone, but where mussel colonistswere notably less abundant than in the vestimentiferanzone. At Worm Hole, where adult mussels were smalland sparse, mussel colonization in the bivalve/suspen-sion zone was not distinctly different from that in thevestimentiferan zone. Finally, colonists of the signaturespecies for the periphery zone, A. pacifica, were moreabundant in the suspension-feeder zones than the pe-riphery (Fig. 6).

In the short-term experiment, colonist abundancesvaried significantly among sites (Table 2). Abundancesof mobile individuals were highest at Worm Hole (Fig.3), where this group was dominated by the lepetodrilidgastropods (Fig. 4). Abundances of sessile individualsalso were highest at Worm Hole, where the ciliate Me-tafolliculina sp. was most numerous. Species abundantat any one site (i.e., those with mean abundances great-er than five individuals per block) were widespreadover all three sites (Table 1).

Temporal variation in environmentand resident biota

Over the course of the colonization studies, therewas little detectable trend in the thermal habitat mea-sured by ‘‘Hobo’’ probes at any site, and (with onenotable exception) little change in the resident faunas.During the 13-mo experiments, tidal variation in tem-perature on daily and fortnightly time scales was evi-dent in the vestimentiferan zones of each site (Mulli-neaux et al. 2000). However, for periods longer than 1mo, mean temperatures remained consistent over theinterval with the exception of the final six weeks atBiovent, where temperatures increased by roughly 28C(Fig. 3 in Mullineaux et al. 2000). Temperatures in thebivalve zones also displayed no longer-term trends(Fig. 7), despite the distinct fluctuations on time scalesof days to weeks, and no temperature anomalies ortrends were detected by ‘‘Hobo’’ probes in the sus-pension-feeder zones (data not shown; temperatureanomalies of ,0.18C were below the detection limitsof ‘‘Hobo’’ probes, and anomalies .0.18C were de-tected with the Alvin probe at only three of 24 blocksin this zone). No change in relative dominance of res-ident species within zones, or movement in zonalboundaries of the resident fauna was detected in re-peated video surveys at any of the sites over the 13-mo study (C. H. Peterson, unpublished manuscript; C.Fisher, unpublished data).

Characterization of temperature variation during thelast 29 mo of the longer-term study was compromisedby the loss of the ‘‘Hobo’’ temperature probe for thatinterval in the vestimentiferan zone, and the misplace-ment of the ‘‘Hobo’’ probe in the bivalve zone into amicrohabitat with no temperature anomaly. Thus, theonly record of temperature change at East Wall for thisinterval was from point measurements by the Alvinprobe during placement and recovery of colonizationblocks. A summary of these sparse measurements from

the vestimentiferan and bivalve zones (Table 4) showslarge variation but suggests no consistent decline orincrease of vent fluid flow during this time. However,in May 1998 we noticed that some areas of the vesti-mentiferan zone that had been vigorously venting inDecember 1995 (as evidenced by shimmering water)had cooled to ambient temperatures.

Despite the absence of a detectable change in ventfluid flows at East Wall, the distribution of adult mus-sels expanded dramatically during the 29-mo intervalbetween December 1995 and May 1998, with individ-uals appearing in the vestimentiferan zone in substan-tial numbers. Thus, by the end of the present study, thefauna inhabiting vigorous vent flow at East Wall be-came similar to the fauna in the comparable habitat atBiovent, where mussels also coexisted with vestimen-tiferans.

DISCUSSION

Negative and positive interactions along a gradientin hydrothermal fluid flux

The null model for our study was an expectation thatearly colonists would have no effect on later ones, lead-ing to no difference in colonization between blocksexposed for continuous intervals and the sum of thoseexposed consecutively. Our results demonstrate thatcolonists of the Mobile functional group, which weremore abundant in the Vestimentiferan zone, deviatedsignificantly from this expectation, allowing us to rejectthe null (tolerance) model of succession for that group.Abundances of individuals in the mobile group werehigher in pooled-consecutive than continuous intervals(Table 2), indicating that direct negative (inhibitory)interactions had occurred. These biological interactionsresulted in faunal changes over time, even in the ab-sence of consistent, long-term changes in the thermal(and inferred chemical) environment.

In contrast, colonists of the Sessile functional grouptended to be more abundant in continuous than pooled-consecutive intervals in the Suspension-feeder zonewhere they were most numerous. This pattern was con-sistent across sites and experimental durations, al-though it was not significant (P . 0.05) in the ANOVA.A sign test indicated that such a pattern was unlikelyto occur (P 5 0.065) in the absence of an intervaleffect. Thus, we suggest that the patterns exhibited bythe sessile group in the suspension-feeder zone are in-dicative of a positive (facilitative) influence of earlycolonists on later arrivals. Indeed, the MANOVA onjoint abundances of sessile and mobile functionalgroups showed a significant shift with interval.

Several prominent individual species appear largelyresponsible for inhibitory or facilitative patterns ex-hibited by the functional groups to which they belong.For three mobile species, the limpets Lepetodrilus ele-vatus and Lepetodrilus pustulosus and the polychaeteOphryotrocha akessoni (except at Biovent), coloniza-


TABLE 3. ANOVAs comparing abundance of individuals between the continuous and pooled-consecutive intervals for theshort-term (13-mo) and longer-term (37-mo) experiments.

Source

L. elevatus

F P

L. pustulosus

F P

O. akessoni

F P

Vestimentiferan

F P

Short termIntervalSiteZoneInt 3 ZoneInt 3 SiteZone 3 SiteZone 3 Site 3 IntPlot (Zone)

Error MS

Sum of squares

0.217.28

70.670.160.182.500.851.37

1.32269.6

0.65,0.01,0.01

0.850.840.070.510.26

0.870.99

86.022.820.250.771.170.99

0.66144.61

0.360.39

,0.010.080.790.550.350.45

11.043.87

37.7410.93

2.542.423.620.68

0.3969.01

,0.01†0.03

,0.01,0.01

0.100.070.020.67

4.850.30

89.503.136.631.031.62

10.28

0.85256.81

0.040.74

,0.010.060.010.410.20

,0.01

Longer termIntervalZoneInt 3 ZonePlot (Zone)

Error MS

Sum of squares

20.5210.68

5.131.92

0.8071.22

,0.01‡0.010.040.22

266.25113.17113.17

0.75

0.0332.25

0.00§0.000.000.66

28.6622.4510.79

5.47

0.2239.00

,0.01\,0.01

0.010.03

15.4147.2715.02

6.00

0.3999.43

0.01¶,0.01,0.01

0.02

Notes: For the short-term experiment, three-way ANOVA included interval, zone (three levels; vestimentiferan, suspension-feeder, and periphery), and site (three levels; Biovent, East Wall, and Worm Hole). In the longer-term experiment, two-wayANOVA included interval and zone (four levels; vestimentiferan, bivalve, suspension-feeder, and periphery). Degrees offreedom as in Table 2. N 5 3 for most cases; except as displayed in Fig. 3. Logarithmic transformations ln(x 1 1) wereperformed on data when necessary to obtain homogeneity of variances. For tests with a significant interval effect (P , 0.05),post hoc Tukey tests shows means that are not significantly different (P . 0.05) connected with underlines (sites are designatedas B, E, and W; zones as v, s, b, and p; intervals as C and P).

† Tukey test: EvP WvP BvC WsP EsP WvC BvP WsC EvC EpP EpC EsC BsP WpP WpC BsC BpC BpP‡ Tukey test: vP bP bC vC pP sC sP pC§ Tukey test: vP bP vC bC sC sP pP pC\ Tukey test: vP bP sC vC sP bC pP pC¶ Tukey test: bP vC vP bC sC sP pC pP

tion in pooled-consecutive intervals was consistentlygreater than in continuous intervals in the vestimen-tiferan zone (Fig. 4). In the sessile protozoan Meta-folliculina sp., the opposite pattern was apparent, withcolonization greater in continuous than pooled-consec-utive intervals in the suspension-feeder zone. Althoughthese interval patterns were statistically significant inthe ANOVAs only for O. akessoni in the short-termexperiments at East Wall and Worm Hole, the sign teston the other three species indicates that the patternswere unlikely to occur (P 5 0.065) in the absence ofan interval effect. Statistical analyses at the specieslevel were less powerful than at the functional grouplevel, but these four species were likely the main con-tributors to the group responses.

The strength of inhibitory and facilitative interac-tions varied along a gradient of hydrothermal fluid fluxat each site, which corresponded to a gradient in faunalcover and metazoan density on the colonization blocks.Inhibitory interactions among the mobile group wereparticularly evident in the vigorous fluid flux zone (ves-timentiferan), where individuals were most abundant(Fig. 3). Facilitative interactions were evident amongthe sessile group only in the weak fluid flux (suspen-sion-feeder) zone. These patterns suggest that com-

petition may have been especially important in the ves-timentiferan zone where metazoan densities were high,whereas facilitation was more prevalent in the suspen-sion-feeder zone where densities were lower. Facili-tation for the sessile group in the suspension-feederzone was barely evident in the longer-term experimentat East Wall. In that case, the abundance pattern of thesessile group was driven largely by patterns of the nu-merically dominant Metafolliculina sp., whereas othersessile taxa in the group (e.g., Bathymodiolus ther-mophilus, Laminatubus alvini, and A. pacifica) showedthe more typical trend of facilitation. The lack of de-tectable facilitation in Metafolliculina in the longer-term exposure suggests that, for this species, processesoperating on periods of years may obliterate the pat-terns generated by facilitation during the early monthsof colonization.

The quantitative demonstration of inhibition of mo-bile species and facilitation of sessile ones in multiplevent sites that differed in successional stages suggeststhat these patterns can be generalized beyond a singletype of vent ecosystem. Similarly, the consistency inthe patterns among species within each functional mo-bility group allows us to predict that similar interac-tions will occur in other vent systems inhabited by these


TABLE 3. Extended.

A. galapagensis

F P

B. thermophilus

F P

L. alvini

F P

A. pacifica

F P

Metafolliculina

F P

0.2120.4388.16

2.510.01

12.802.441.43

0.53169.728

0.65,0.01,0.01

0.100.99

,0.010.070.24

0.206.62

26.897.270.116.173.822.18

0.4064.27

0.660.01

,0.01,0.01

0.900.000.010.08

2.058.96

30.091.942.432.640.851.46

0.99136.2

0.16,0.01,0.01

0.160.110.060.510.23

0.5915.9154.52

0.580.551.230.560.28

1.10196.8

0.45,0.01,0.01

0.570.590.320.690.94

1.3711.4825.53

0.432.401.060.391.16

4.78572.2

0.25,0.01,0.01

0.660.110.400.820.36

0.6870.68

0.785.00

0.2975.62

0.44,0.01

0.550.03

0.1717.45

0.201.96

0.5843.67

0.70,0.01

0.890.21

0.088.090.261.18

1.3554.60

0.790.020.850.43

4.8093.25

0.5015.38

0.25105.30

0.07,0.01

0.70,0.01

1.0421.44

1.011.11

2.47205.97

0.35,0.01

0.450.46

FIG. 7. Temperatures recorded in the bivalve zone over a 5-mo interval (November 1994–April 1995) at Biovent andover an 8-mo interval (April 1995–December 1995) at East Wall. Records for the 5-mo interval at East Wall are not shownbecause the temperature probe was not recovered. Records for the 5-mo interval at Worm Hole and the 8-mo intervals atBiovent and Worm Hole are not shown because no temperature anomaly was recorded at the probe locations.

functional groups, even if they do not host the samespecies. However, the difference in the relativestrengths of inhibitory and facilitative interactionsalong a gradient in vent fluid flux (and inferred gradientin productivity) suggests that predicting which inter-action will dominate at a specific location requiresknowledge of the local hydrothermal fluid flux. Ourability to generalize and predict from these deep-seaexperiments demonstrates the value of conducting ma-nipulative studies over a broad range of environmentalconditions, even in environments where such approach-es are especially challenging.

Mechanisms for succession along a gradientin vent fluid flux

Any mechanisms proposed to account for the ob-served biological interactions must explain both the

difference between responses of mobile and sessilefunctional groups to initial colonists, and the differencein the intensity and direction of these patterns along agradient in vent fluid flux, productivity and faunal den-sity. We suggest that the following processes contributeto successional changes in the vent community: (1)mobile species exert negative influences on both mo-bile and sessile colonists by removing (‘‘bulldozing’’)new settlers; (2) sessile species are relatively immuneto removal once firmly attached and relatively large;(3) sessile species exert negative influences on mobilecolonists by occupying their grazing space; and (4)sessile species exert positive effects on sessile colonistsby providing spatial refuge from bulldozing by mobilegrazers, and/or a settlement cue. We expect negativeinteractions to be more prominent in regions of high


TABLE 4. Temperature anomalies recorded at the base of blocks deployed at East Wall inApril and December 1995 and recovered May 1998 (37-mo and 29-mo intervals).

ZoneInterval

(mo)Deploy

temp. (8C)Recover

temp. (8C)Change

temp. (8C)

Vestimentiferan 29292937

1.54.12.05.5

5.41.11.51.1

3.923.020.524.4

Bivalve 293737

2.95.31.0

0.06.60.0

22.91.3

21.0

Note: Only blocks with temperatures recorded successfully on deployment and recovery areshown.

faunal density where disruption by mobile grazers isfrequent and grazing space limited. Positive interac-tions are more likely to occur in regions of lower faunaldensity where bulldozing is less frequent, allowing ses-sile species to become established and more effectivelyprovide a refuge or settlement cue to subsequent sessilecolonists. A diagram of these processes (Fig. 8) showsthe relative importance of negative interactions de-creasing and positive interactions increasing along agradient of decreasing vent fluid flux and metazoanfaunal density.

Support for these proposed interactions among andbetween mobile and sessile species comes from a mech-anistic understanding of faunal interactions in othermarine communities and from natural history obser-vations of vent species. Mobile grazers such as limpetsare known to impede attachment of other mobile orsessile species by physically dislodging (bulldozing)their recruits (Connell 1961, Stimson 1970, Dayton1971, Menge 1976, Dixon 1978) and potentially caninterfere with each other’s feeding activities. Similarly,the presence of established sessile organisms mayphysically limit the size of, or access to, grazing areasof mobile individuals (note that although large vesti-mentiferan tubes can ultimately provide grazing areafor some limpet species, tubes remained small in thepresent study). Their influence on the physical habitatis analogous in some ways to the effects of plants interrestrial systems as ‘‘ecosystem engineers’’ (Jones etal. 1997). When new habitat is exposed at vents, thesmall mobile grazers have the potential to impede es-tablishment of sessile species by removing newly set-tled individuals. However, it appears that at least somesessile individuals are able to recruit, and, once theygrow to sufficient size, may begin to restrict the grazingterritories and abilities of mobile species, thereby lim-iting the number of mobile colonists and causing theinhibition detected in our experiments. Alternatively,at high densities, mobile grazers also may removeenough food (expected to be surface-attached mi-crobes; Hessler and Smithey 1983) to limit the foodavailable for other grazing individuals. Thus, the in-hibition of mobile individuals in the vestimentiferanzone could be due to direct interference from estab-

lished mobile or sessile colonists, and/or resource lim-itation among the mobile species.

Recruitment of sessile individuals may be facilitatedby established sessile species through two differentmechanisms. Gregarious settlement mediated by chem-ical or physical cues produced by the adults is commonamong tubiculous species (reviewed in Pawlik 1992),suggesting a potential mechanism for facilitative re-cruitment of sessile vent species (Mullineaux et al.2000). Alternatively, established sessile species couldfacilitate other sessile individuals by interfering withgrazing activities of mobile species and providing arefuge from bulldozing for new sessile recruits. Oneway to test for gregarious settlement on our coloni-zation blocks would be to look for small-scale (milli-meters to centimeters) aggregations in the colonists.However, such an analysis was outside the scope ofour study, and thus we cannot use our colonization datato evaluate directly which process (provision of settle-ment cues or refuge from bulldozing) provides the mostplausible explanation for the observed facilitation ofsessile species. Although facilitation via a refuge frombulldozing may be important where mobile grazers areabundant, facilitation via a settlement cue seems a morelikely explanation for facilitation in the suspension-feeder zone where mobile grazers are rare.

The prevalence of inhibitory interactions at the highend of the fluid flux gradient, where productivity andfaunal densities are high, and of facilitative interactionsat the low end, where productivity and densities arelow, is consistent with our proposed mechanisms ofspecies interactions. We expect negative interactions tobe more intense in high-productivity regions, wherefaunal densities are high and mobile individuals canfrequently bulldoze subsequent mobile settlers, or limittheir foraging efficiency. Interference of mobile colo-nists by sessile individuals also is more likely at highfaunal densities, where space is limited and sessile spe-cies not easily avoided. However, inhibition of sessilespecies in the high fluid flux region was not detected,suggesting that sessile settlers were less susceptible tobulldozing than mobile ones and/or that facilitation byestablished sessile individuals (via chemical cues orrefuges from bulldozing) more than compensated for


FIG. 8. Diagram showing proposed mechanisms for species interactions along a production gradient at hydrothermalvents. Wedges indicate the relative intensity of hydrothermal fluid flux (as inferred from temperature measurements), andnegative interactions and positive interactions (as indicated by interval effects in colonization experiments). Zonation ofresident fauna indicates position of signature species for the vestimentiferan zone (vestimentiferans, including Riftia pachyptila,Tevnia jerichonana, and Oasisia alvinae), bivalve zone (vent mussel, Bathymodiolus thermophilus), suspension-feeder zone(serpulid polychaete, Laminatubus alvini), and periphery zone (foraminifer, Abyssotherma pacifica) along the fluid fluxgradient. Supplementary evidence from vents and other environments suggests that facilitation of sessile species may be dueto settlement cues and/or refuges from disturbance, whereas inhibition of mobile species may be due to disruption andcompetition from other mobile individuals, impediments to grazing imposed by sessile individuals, and/or indirect effects ofsize-selective consumption by fish. Dotted lines in the periphery zone indicate absence of detectable vent fluid flux anduncertainty about biological interactions due to small numbers of colonists.

interference by mobile ones. We expect positive inter-actions to be most prominent at the low end of the fluidflux gradient, where densities of mobile species arelow, and their potential to interfere with sessile settlersminimal. In this region, facilitation of sessile settlersby established sessile individuals (most likely viachemical or physical settlement cues) should be readilydetectable because it is not negated by inhibitory in-teractions. Furthermore, in the regions of weak ventfluid flux, vent fluid chemical and thermal cues arediluted, and biogenic cues may provide the best strat-egy for avoiding erroneous settlement into the food-poor suspension-feeder and periphery zones (Mulli-neaux et al. 2000).

Indirect effects on succession

Although our discussion of species interactions hasconcentrated mainly on direct effects of invertebratecolonists on each other, successional changes alsocould have been mediated by mobile predators that alterabundances of colonizing species. A series of inde-pendent observations on the behavior of the vent fishThermarces cerberus, which is an abundant predatorat vents along the northern East Pacific Rise (Geist-doerfer 1996), suggests that it potentially has a stronginfluence on the population density and size structureof limpets at vents. Gut contents show that a prominentcomponent of this fish’s diet are limpets, including Le-petodrilus spp. (Geistdoerfer 1996). Caging experi-


ments have demonstrated that gastropods attain higherdensities when T. cerberus has been excluded (Micheliet al. 2002). Recent feeding observations indicate thatT. cerberus concentrates its feeding on larger limpets,rather than the smaller size classes that are early col-onists of newly cleared habitat (G. Sancho, unpublisheddata). Considered together, these observations suggestthat T. cerberus may reduce the population sizes ofLepetodrilus spp. and other limpets at vents, but in asize-selective manner that gives the small, newly re-cruited individuals a size refuge.

Given these observations on the trophic role of Ther-marces cerberus, we suggest that its predatory activ-ities may be contributing to the inhibition pattern ob-served in mobile species in the present study (Fig. 8).When new habitat becomes available, mobile speciescolonize and appear to be immune to fish predationwhile they are small. Once the limpets reach a certainsize, they become prey for the fish T. cerberus, and arecropped to lower levels. This process would result inlower limpet abundance in continuous experimental ex-posures where limpets reach larger sizes than in pooled-consecutive exposures. The effect of the vent fish isexpected to be important in the vestimentiferan zone,but not in the other zones, where they rarely are ob-served feeding (L. Mullineaux, C. Petersen, F. Micheli,and S. Mills, personal observations) and have littleeffect on limpet populations (Micheli et al. 2002).

Microhabitat variation

Any interpretation of interval and zone effects oncolonization must take into account the substantialthermal and chemical variation within zones, which canoccur over scales as small as a few centimeters (John-son et al. 1988). In the present study, this variation wasdocumented in the form of water temperature mea-surements at individual blocks (complete data notshown). For instance, at East Wall, block temperaturesdiffered from each other by as much as 22.78C withinthe vestimentiferan zone, 14.48C within the bivalvezone, and 0.98C within the suspension-feeder zone (noanomaly or variation was detected in any of the pe-riphery zones). This variation in vent fluid flow onsmall spatial scales (centimeters to meters) is likelyresponsible, in large part, for the observed high vari-ability in colonization among blocks within a treatment(interval or zone) and the occasionally significant var-iations among plots within zones (e.g., for the sessilegroup in the longer-term experiment; Table 2). Micro-habitat variation does not introduce bias into our anal-yses because variation among zones was much greaterthan variation within zones (Table 2), but it does il-lustrate that although our zones appear relatively ho-mogeneous in terms of dominance by prominent sig-nature species, they are not homogeneous environ-mentally.

The substantial variation in vent fluid flux over verysmall spatial scales was a surprise to us, and compro-

mised our experimental design. A split plot designclearly is not ideal in a habitat where environmentalvariability is high within plots. Furthermore, to in-crease power of experimental tests in such systems, thelevel of replication needs to exceed three (substantial-ly) or a suite of temperature recorders must be em-ployed to provide temperature covariates for each ex-perimental unit to reduce error variance. Such exper-imentation is becoming possible in the deep sea asmanipulative, observational and sampling capabilitiesimprove, and background information on the naturalhistory and dynamics of deep ecosystems becomesavailable.

Interpreting prior successional observations at vents

Vestimentiferan colonization patterns were not stud-ied at the species level in the present study, but ourresults do contribute to a better understanding of themechanisms causing the successional transition fromvestimentiferans to bivalves observed in prior studies(Hessler et al. 1988, Shank et al. 1998). The musselBathymodiolus thermophilus colonized new habitat inthe vigorous flow (vestimentiferan) zone quickly (in#5 mo) and in substantially higher numbers than inthe other zones (Fig. 5). This result is surprising be-cause the mussel is the signature adult species of themoderate fluid flux (bivalve) zone, and has not beenobserved as a pioneer colonist in photo- and video-based analyses of succession in the vigorous fluid fluxzones of other vent sites along the East Pacific Rise(Hessler et al. 1988, Shank et al. 1998). One scenarioconsistent with all these observations is that musselsare early colonists of the Vestimentiferan zone but re-main cryptic until they reach an age of .13 mo. An-other possibility is that mussels are facilitated by thestructure provided by mature vestimentiferans, perhapsvia the increased surface area provided by their tubes.The impact of the mussels on the much larger vesti-mentiferans is unlikely to be substantial until they reacha size or population density that can divert flow fromthe vestimentiferan plumes (as proposed in Johnson etal. 1994) or otherwise disrupt them mechanically.Thus, we suggest that the transition from vestimenti-ferans to mussels at EPR vent sites involves facilitationof mussels by the sessile pioneer, which is subsequentlyoutcompeted when the mussels reach a critical size ordensity.

Are vent communities controlled by biologicalor physical processes?

In summary, temporal change in colonization pat-terns of species at deep-sea hydrothermal vents appearsto reflect inhibition of mobile species by initial colo-nists (both mobile and sessile), and possibly facilitationof sessile species by initial sessile colonists. The rel-ative intensity of negative and positive interactionsvaries along a production gradient, with negative in-teractions most prominent where vent fluid flux and


faunal densities are highest. Our results suggest thatbiological interactions at the functional (mobility)group level can shape succession in a general, repeat-able sequence, in the absence of consistent environ-mental change. Although physiological constraints andnutritional requirements of individual species no doubtset limits to their occurrence in the patchy and change-able vent environment (Fisher et al. 1988a, b), we con-clude that their interactions with each other and withmobile predators have a strong influence on succession.

The development of our understanding of commu-nity structure at hydrothermal vent habitats reveals aparallel with ecological study in rocky intertidal hab-itats. Both systems exhibit strong gradients in the phys-ical or chemical environment, and are characterized byfaunal zonation that correlates with the gradients. Inthe rocky intertidal, manipulative studies have dem-onstrated that biological interactions are important inthe dynamics and community structure of the ecosys-tem (reviewed in Connell 1972, Paine 1977). Com-parable studies in deep-sea vent communities have thepotential to produce a similarly enhanced perspectiveon the complex set of biological processes involved inestablishing zonation patterns during community suc-cession. As remote underwater technologies are de-veloped, and ecological questions focused, it will be-come possible to design experiments with sufficientreplication to test our proposed successional mecha-nisms and resolve current controversies over the rel-ative importance of physical and biological processesat vents.

ACKNOWLEDGMENTS

We thank the captains, crews, and Alvin Group memberson four separate cruises on the Atlantis II and Atlantis forsupport and assistance beyond the call of duty. Chuck Fisherwas involved in all stages of this project, providing ideas,labor, natural-history information, and enthusiasm. Scientificassistance at sea was generously provided by Ewann Bernt-son, Galen Johnson, Hunter Lenihan, Heather Hunt, StacyKim, Anna Metaxas, Gorka Sancho, and Shannon Smith. DanFornari provided invaluable maps of seafloor bathymetry andhydrothermal sites. Prior drafts of this paper were improvedby the comments of Stace Beaulieu, Lara Gulmann, RobertJennings, and Tim Shank. Vicke Starczak assisted with thestatistical analyses. This work was supported by the NationalScience Foundation grants OCE-9315554 and OCE-9712233to L. Mullineaux, OCE-9317735 and OCE-9712809 to C.Peterson, and OCD-9317737 and OCD-9712808 to C. R.Fisher. Parts of this study were conducted while F. Micheliwas a postdoctoral researcher supported by the NSF NationalCenter for Ecological Analysis and Synthesis, the Universityof California Santa Barbara, and the California ResourceAgency. This is WHOI contribution 10943.

LITERATURE CITED

Ambrose, W. G. J. 1984. Influence of residents on the de-velopment of a marine soft-bottom community. Journal ofMarine Research 42:633–654.

Benedetti-Cecchi, L. 2000. Predicting direct and indirect in-teractions during succession in a mid-littoral rocky shoreassemblage. Ecological Monographs 70:45–72.

Berlow, E. 1997. From canalization to contingency: historicaleffects in a successional rocky intertidal community. Eco-logical Monographs 67:435–460.

Butterfield, D., I. Jonasson, G. Massoth, R. Feely, K. Roe,R. Embley, J. Holden, R. McDuff, M. Lilley, and J. De-laney. 1997. Seafloor eruptions and evolution of hydro-thermal fluid chemistry. Philosophical Transactions of theRoyal Society of London, Series A 355:369–386.

Callaway, R. M., and L. R. Walker. 1997. Competition andfacilitation: a synthetic approach to interactions in plantcommunities. Ecology 78:1958–1965.

Caswell, H. 1976. Community structure: a neutral modelanalysis. Ecological Monographs 46:327–354.

Clements, F. E. 1916. Plant succession: an analysis of thedevelopment of vegetation. Carnegie Institution of Wash-ington, Washington, D.C., USA.

Connell, J. H. 1961. Effects of competition, predation byThais lapillus, and other factors on natural populations ofthe barnacle Balanus balanoides. Ecological Monographs31:61–104.

Connell, J. H. 1972. Community interactions on marine rockyintertidal shores. Annual Reviews of Ecology and System-atics 3:169–172.

Connell, J. H., and R. O. Slatyer. 1977. Mechanisms of suc-cession in natural communities and their role in communitystability and organization. American Naturalist 111:1119–1144.

Dayton, P. K. 1971. Competition, disturbance, and commu-nity organization: the provision and subsequent utilizationof space in a rocky intertidal community. Ecological Mono-graphs 41:351–389.

Dixon, J. D. 1978. Determinants of the local distribution offour closely-related species of herbivorous marine snails.Dissertation. University of California, Santa Barbara, Cal-ifornia, USA.

Egler, F. E. 1954. Vegetation science concepts I. Initial flo-ristic composition. A factor in old-field vegetation devel-opment. Vegetatio 4:412–417.

Farrell, T. 1991. Models and mechanisms of succession: anexample from a rocky intertidal community. EcologicalMonographs 61:95–113.

Fisher, C. R., et al. 1988a. Microhabitat variation in thehydrothermal vent mussel Bathymodiolus thermophilus, atRose Garden vent on the Galapagos rift. Deep-Sea Research35:1769–1792.

Fisher, C. R., et al. 1988b. Microhabitat variation in thehydrothermal vent clam, Calyptogena magnifica, at RoseGarden vent on the Galapagos rift. Deep-Sea Research 35:1811–1832.

Franklin, J., and J. MacMahon. 2000. Messages from a moun-tain. Science 288:1183–1185.

Fustec, A., D. Desbruyeres, and S. K. Juniper. 1987. Deep-sea hydrothermal vent communities at 138N on the EastPacific Rise: Microdistribution and temporal variations. Bi-ological Oceanography 4:121–164.

Geistdoerfer, P. 1996. L’ichthyofaune des ecosystemes as-socies a l’hydrothermalisme oceanique: etat des connaiss-ances et resultats nouveaux. Oceanologica Acta 19:539–548.

Grassle, J. F., and L. S. Morse-Porteous. 1987. Macrofaunalcolonization of disturbed deep-sea environments and thestructure of deep-sea benthic communities. Deep-Sea Re-search 34:1911–1950.

Haymon, R. M., et al. 1993. Volcanic eruption of the mid-ocean ridge along the East Pacific Rise crest at 98 45–529N:direct submersible observations of sea-floor phenomena as-sociated with an eruption event in April, 1991. Earth Plan-etary Science Letters 119:85–101.

Hessler, R. R., and W. M. Smithey. 1983. The distributionand community structure of megafauna at the Galapagos


Rift hydrothermal vents. Pages 735–770 in P. S. Rona, K.Bostrom, L. Laubier, and K. L. Smith, editors. Hydrother-mal processes at seafloor spreading centers. NATO Con-ference Series. Plenum Press, New York, New York, USA.

Hessler, R. R., W. M. Smithey, M. A. Boudrias, C. H. Keller,R. A. Lutz, and J. J. Childress. 1988. Temporal change inmegafauna at the Rose Garden hydrothermal vent (Gala-pagos Rift; eastern tropical Pacific). Deep-Sea Research 35:1681–1709.

Hessler, R. R., W. M. Smithey, and C. H. Keller. 1985. Spatialand temporal variation of giant clams, tubeworms and mus-sels at deep-sea hydrothermal vents. Bulletin of the Bio-logical Society of Washington 6:465–474.

Hixon, M. A., and W. N. Brostoff. 1996. Succession andherbivory: effects of differential fish grazing on Hawaiiancoral-reef algae. Ecological Monographs 66:67–90.

Johnson, K. S., J. J. Childress, C. L. Beehler, and C. M.Sakamoto. 1994. Biogeochemistry of hydrothermal ventmussel communities: the deep-sea analogue to the intertidalzone. Deep-Sea Research 41.

Johnson, K. S., J. J. Childress, R. R. Hessler, C. M. Sakamoto-Arnold, and C. L. Beehler. 1988. Chemical and biologicalinteractions in the Rose Garden hydrothermal vent field,Galapagos spreading center. Deep-Sea Research 35:1723–1744.

Jones, C. G., J. H. Lawton, and M. Shachak. 1997. Positiveand negative effects of organisms as physical ecosystemengineers. Ecology 78:1946–1957.

Lubchenco, J. 1983. Littorina and Fucus: effects of herbi-vores, substratum heterogeneity, and plant escapes duringsuccession. Ecology 64:1116–1123.

MacDonald, K. C., K. Becker, F. N. Spiess, and R. D. Ballard.1980. Hydrothermal heat flux of the ‘‘black smoker’’ ventson the East Pacific Rise. Earth and Planetary Science Let-ters 48:1–7.

McCook, L. J. 1994. Understanding ecological communitysuccession: causal models and theories, a review. Vegetatio110:115–147.

Menge, B. A. 1976. Organization of the New England rockyintertidal community: role of predation, competition, andenvironmental heterogeneity. Ecological Monographs 46:355–393.

Micheli, F., C. H. Peterson, L. S. Mullineaux, C. R. Fisher,S. W. Mills, G. Sancho, G. A. Johnson, and H. S. Lenihan.2002. Predation structures communities at deep-sea hy-drothermal vents. Ecological Monographs 72:365–382.

Mullineaux, L. S., C. R. Fisher, C. H. Peterson, and S. W.Schaeffer. 2000. Vestimentiferan tubeworm succession at

hydrothermal vents: use of biogenic cues to reduce habitatselection error? Oecologia 123:275–284.

Odum, E. P. 1969. The strategy of ecosystem development.Science 164:262–270.

Paine, R. T. 1977. Controlled manipulations in the marineintertidal zone, and their contributions to ecological theory.Pages 245–270 in The changing scenes in natural sciences,1776–1976. Academy of Natural Sciences, Philadelphia,Pennsylvania, USA.

Pawlik, J. 1992. Chemical ecology of the settlement of ben-thic marine invertebrates. Oceanography and Marine Bi-ology: an Annual Review 30:273–335.

Shank, T. M., D. J. Fornari, K. L. von Damm, M. D. Lilley,R. M. Haymon, and R. A. Lutz. 1998. Temporal and spatialpatterns of biological community development at nascentdeep-sea hydrothermal vents (98509 N, East Pacific Rise).Deep-Sea Research 45:465.

Sokal, R. R., and F. J. Rohlf. 1995. Biometry. W. H. Freeman,New York, New York, USA.

Sousa, W. P. 1979a. Disturbance in marine intertidal boulderfields: the non-equilibrium maintenance of species diver-sity. Ecology 60:1225–1239.

Sousa, W. T. 1979b. Experimental investigations of distur-bance and ecological succession in a rocky intertidal algalcommunity. Ecological Monographs 49:227–254.

Stimson, J. 1970. Territorial behavior of the owl limpet, Lot-tia gigantea. Ecology 51:113–118.

Sutherland, J. P., and R. H. Karlson. 1977. Development andstability of the fouling community at Beaufort, North Car-olina. Ecological Monographs 47:425–446.

Underwood, A. J., and E. J. Denley. 1984. Paradigms, ex-planations, and generalizations in models for the structureof intertidal communities on rocky shores. Pages 151–180in D. R. Strong, D. Simberloff, L. G. Abele, and A. B.Thistle, editors. Ecological communities: conceptual issuesand the evidence. Princeton University Press, Princeton,New Jersey, USA.

von Damm, K. L. 1995. Controls on the chemistry and tem-poral variability of seafloor hydrothermal fluids. Pages222–247 in S. E. Humphris, R. A. Zierenberg, L. S. Mul-lineaux, and R. E. Thomson, editors. Seafloor hydrothermalsystems: physical, chemical, biological, and geological in-teractions. Geophysical Monograph Series. American Geo-physical Union, Washington, D.C., USA.

von Damm, K. L. 2000. Chemistry of hydrothermal ventfluids from 98–108 N, East Pacific Rise; ‘‘time zero,’’ theimmediate posteruptive period. Journal of Geophysical Re-search, B, Solid Earth and Planets 105:11,203–11,222.

Witman, J. K. 1987. Subtidal coexistence: storms, grazing,mutualism, and the zonation of kelps and mussels. Eco-logical Monographs 57:167–187.

successional mechanism varies along a gradient in ...hydrothermal fluid flux at deep-sea vents...

Documents