Estuarine,CoastalandShelfScience200(2018)1e18 ContentslistsavailableatScienceDirect Estuarine, Coastal and Shelf Science journal homepage: www.elsevier.com/locate/ecss Interplay between abiotic factors and species assemblages mediated by the ecosystem engineer Sabellaria alveolata (Annelida: Polychaeta) Auriane G. Jones a,b,c,*, Stanislas F. Dubois a, Nicolas Desroyb, Je(cid:1)ro^me Fournierc,d aIFREMER,CentredeBretagne,DYNECOLEBCO,29280Plouzan(cid:1)e,France bIFREMER,EnvironnementetRessourcesBretagnenord,38rueduPortBlanc,BP80108,35801Dinardcedex,France cCNRS,UMR7208BOREA,61rueBuffon,CP53,75231Pariscedex05,France dMNHN,StationdeBiologieMarine,BP225,29182Concarneaucedex,France a r t i c l e i n f o a b s t r a c t Articlehistory: Sabellaria alveolata is a gregarious polychaete that uses sand particles to build three-dimensional Received1February2017 structuresknownasreefs,fixedatoprocksorbuiltonsoftsediments.Thesestructuresareknownto Receivedinrevisedform modifythelocalgrain-sizedistributionandtohostahighlydiversifiedmacrofauna,alteredwhenthe 4October2017 reefundergoesdisturbances.Thegoalofthisstudywastoinvestigatethedifferentsedimentaryand Accepted6October2017 biologicalchangesassociatedwiththepresenceofaS.alveolatareefovertwocontrastingseasons(late Availableonline7October2017 winter and late summer), and how these changes were linked. Three different sediments were considered: the engineered sediment (the actual reef), the associated sediment (the soft sediment Keywords: surroundingthereefstructures)andacontrolsoftsediment(i.e.noreefstructuresincloseproximity). Honeycombworm Macrobenthos Univariate and multivariate comparisons of grain-size distribution, soft sediment characteristics Benthicprimaryproduction (organic matter content, chlorophyll a, pheopigments and carbohydrate concentrations) and macro- Habitatdisturbance fauna were conducted between the different sediment types at both seasons and between the two Silt seasonsforeachsedimenttype.Adistance-basedredundancyanalyses(dbRDA)wasusedtoinvestigate Betadiversity thelinkbetweenthedifferentenvironmentalparametersandthemacrofaunaassemblages.Finally,we France focusedonadisturbancecontinuumoftheengineeredsedimentsproxiedbyanincreaseinthemud Brittany presentinthesesediments.Theeffectsofacontinuousandincreasingdisturbanceontheassociated MontSaint-MichelBay faunawereinvestigatedusingpairwisebetadiversityindices(SørensenandBray-Curtisdissimilarities andtheirdecompositionintoturnoverandnestedness).Resultsshowedasignificanteffectofthereef on the local sediment distribution (coarser sediments compared to the control) and on the benthic primary production (higher in the associated sediments). At both seasons, S. alveolata biomass and sedimentprincipalmodeweretheenvironmentalparameterswhichbestdifferentiatedtheengineered, associated and control sediment assemblages. These two parameters are under the ecosystem engi- neer's influence stressing its importance in structuring benthic macrofauna. Furthermore, in late summerbutnotinlatewinter,presence/absenceandabundance-basedbetadiversitywerepositively correlatedtoourdisturbanceproxy(mudcontent)atendencydrivenbyaspeciesreplacementanda riseintheassociatedfaunadensity.OurfirstsetofresultshighlighttheimportanceofS.alveolatareefs asbenthicprimaryproductionenhancersviatheirphysicalstructureandtheirbiologicalactivity.The resultsobtainedusingbetadiversityindicesemphasizetheimportanceofrecruitmentinstructuring the reef's macrofauna and e paradoxically e the ecological value of S. alveolata degraded forms as biodiversityandrecruitmentpromoters. ©2017ElsevierLtd.Allrightsreserved. Abbreviations:MSMB,MontSaint-MichelBay;MPB,Microphytobenthos;TOM, 1. Introduction Total organic matter; Chl a, Chlorophyll a; Pheo, Pheopigments; Ins, Insoluble carbohydrates; Sol, Soluble carbohydrates; dbRDA, Distance-based redundancy Ecosystemengineersareorganismscapableofmodifyingtheir analysis;CPUE,Catch-per-unit-effort. * Corresponding author. IFREMER, Centre de Bretagne, DYNECO, Laboratoire local environment through their physical presence (i.e. autogenic d’Ecologie Benthique Co^tie(cid:3)re (LEBCO), ZI Pointe du Diable, CS 10070, 29280 engineers)and/ortheirbiologicalactivity(i.e.allogenicengineers), Plouzane(cid:1),France. “directly or indirectly modulating the availability of resources to E-mailaddress:[email protected](A.G.Jones). https://doi.org/10.1016/j.ecss.2017.10.001 0272-7714/©2017ElsevierLtd.Allrightsreserved. 2 A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 otherspecies”(Jonesetal.,1994).Ultimately,thesespeciesmain- theEuropeancoast(i.e.MSMBandBourgneufBayinFrance)(Holt tain,modify,createorevendestroyhabitats(Boumaetal., 2009; et al., 1998). Nonetheless, they constitute exceptional locations Jonesetal.,1994).Theabioticmodificationscausedbyecosystem composed of two distinct entities: the actual three-dimensional engineerscanleadtofacilitationforsomeorganisms(Hackerand reef structures (engineered sediment), which is spatially discon- Gaines,1997) and inhibitionthrough negativespecies interaction tinuousandthesoftsedimentpresentbetweenthereefstructures for others (Bouma et al., 2009; Jones et al., 1997). Nonetheless, (associated sediment) (Desroy et al., 2011). Several kilometers bioengineered habitats are often reported tohost a morediverse separate them from the nearest rocky shore which signifies, in species assemblage than the adjoining non-engineered habitats contrast to the veneer form of S. alveolata structures, complete (Ataide et al., 2014; De Smet et al., 2015; Jones et al., 1997; isolationfrommostofthejuvenileandadultfaunainhabitingthese Stachowicz, 2001). Physical ecosystem engineering appears to be rocky shores. Furthermore, their physical borders are easy to particularly important where the environment is extreme (e.g. visualizeagainstthesurroundingsoftsediment.Thesesitesgiveus thermic, hydrodynamic and/or hydric stress), like in temperate the chance to study different components of S. alveolata's engi- intertidalareas(Boumaetal.,2009;Jonesetal.,1997).Indeed,ac- neeringeffect(Passarellietal.,2014;Wrightetal.,2006).Thisen- cordingtoJonesetal.(1997,1994),theseextremeconditionsmight gineering effect can be seen from both an environmental and a have favored the selection of “extended phenotype engineers” biological perspective by looking at how the ecosystem engineer through enhanced survival of the engineer and the cohabiting modifiesthelocalsedimentarycharacteristicsandhowthebiodi- fauna (Dawkins, 1982). These engineer species create complex versitychangesbetweenacontrolsediment,theassociatedandthe habitatsthatreducelocalpressuressuchaspredationorthermal engineered sediments. The control soft sediment represents the stress, whilst increasing biodiversity (Bouma et al., 2009). Ulti- baseline or the unmodified state before the honeycomb worms mately, such favorable environmental changes can lead to an start building reefs, hence representing a new structural state interestingparadoxwhere“thespatialextentoftherealizedniche (Jonesetal.,2010). ofaspeciescanbelargerthanthespatialrangepredictedbythe This biogenic habitat is not structurally homogenous, mainly fundamental niche” as described by Bruno et al. (2003) and re- due to multiple disturbances; direct natural disturbances such as portedformusselsandbarnaclesinAscophyllumnodosumcanopies stormsandcoldwinters,directanthropogenicdisturbancessuchas byBertnessetal.(1999). tramplingandindirectanthropogenicdisturbancesthroughshell- Temperate coasts host a striking number of ecosystem engi- fishfarmingandcoastalengineering.Thesedisturbancesleadtoa neeringspecies,spanningfrommollusks(forareviewseeGutie(cid:1)rrez gradual modification of the reef visible through disaggregation, etal.(2003))andpolychaetes(e.g.Laniceconchilega(DeSmetetal., increasingfinesediments,decreasingecosystemengineerdensity 2015)) to canopy-forming algae (e.g. Ascophyllum nodosum andincreasingepibiontcover,causinganumberofchangesinthe (Bertnessetal.,1999)).AlongtheEuropeancoastline,aparticular associatedfauna(Duboisetal.,2006b,2002;Plicantietal.,2016). ecosystem engineer has the ability to build three-dimensional Modifications of the associated fauna have been investigated in structures on top of sediments qualified as reefs (Holt et al., severalcategoricalwaysbutneveralongadisturbancecontinuum 1998).Thisspeciesisacommongregarioustubiculouspolychaete (Duboisetal.,2006b,2002;Plicantietal.,2016).Tounderstandthe calledSabellariaalveolata,a.k.a.thehoneycombworm.Itgenerally changesintheassociatedfaunaalongthiscontinuum,wechoseto livesintheintertidalzonefrommidtolowtidelevelsandcanbe focusonthebetadiversityseenas“theextentofchangeincom- found from Scotland and Ireland to Morocco (Muir et al., 2016). munity composition” as defined by Whittaker (1960) and on an Sabellaria alveolata uses sand particles remobilized bywaves and abundance-baseddissimilaritymeasurementusingtheBray-Curtis tidalactiontobuildthetubeinwhichitlives(LeCametal.,2011). dissimilarity.AnalyzingbetadiversityinaS.alveolatareefcanhelp SincethepelagiclarvaeareattractedbytheL-dopapresentinthe us understand the functioning of this biogenic habitat and give organic cement produced by the adult worms for their tube- more relevant information to decision makers regarding conser- buildingactivity,theywilltendtosettleonexistingreefs(Pawlik, vationissues.First,takingintoaccountthethreepreviouslydefined 1988; Wilson, 1968). This phenomenon coupled with favorable sedimenttypes(control,associatedandengineeredsediments),we environmental conditions(i.e. grain-size structure, hydrodynamic tested in a categorical way, the following hypotheses: (1) the processes,foodavailabilityandwatertemperature)canleadtothe engineered sediment affects the different sedimentarycharacter- development of large biogenic reefs (Holt et al., 1998). These istics of the associated sediment, especially grain-size, organic structures are commonly found on rocky substrata as veneers or mattercontent and microphytobenthos and (2) the diversityand hummockswheretheyrarelyexceed50cminheightforafewtens species composition of both the engineered and the associated ofsquaremetersbutinsomerareinstances,theycanbefoundin sedimentsaredifferentfromthecontrolsediment.Wealsolooked soft bottom areas where theycan grow up to 2 m in height and into potential changes between late winter and late summer, severalhectaresinsize(Holtetal.,1998;Noernbergetal.,2010). regardingsedimentcompositionandmacrofaunaassemblagesfor Thelargestoftheseformations,whichisalsothelargestbiogenic each sediment type. Then, using beta diversity and dissimilarity habitatinEurope,islocatedintheMontSaint-MichelBay(MSMB) measurements,wetestedthefollowinghypothesis:anincreasing inFrance(Desroyetal.,2011;Duboisetal.,2002). disturbanceoftheengineeredsedimentpromotes(1)betadiversity The research around this species has mainly focused on its and more specifically species turnover and (2) abundance-based physiology (i.e. reproduction, fecundity, feeding mode) (Dubois dissimilarityandmorespecificallyabundancegradients. et al., 2003, 2005, 2006a, 2009) and its tube building activity (Fournieretal.,2010;LeCametal.,2011).Otherstudieshavelooked 2. Materialsandmethods intotheecologyofreefswithaparticularinterestontheassociated fauna (Dias and Paula, 2001; Porta and Nicoletti, 2009; Schlund 2.1. Studyarea et al., 2016) and factors influencing it such as the reef's different growthstages(Duboisetal.,2002),epibionts(Duboisetal.,2006b), ThisstudytookplaceinthecentralpartoftheMSMBwherethe human trampling (Plicanti et al., 2016) and ecological status largestbioconstructioninEuropeislocated;theSainte-Annereef (Desroyetal., 2011).A largepartof thesestudies has focused on (48(cid:1)380700N and 1(cid:1)400100W), built by the honeycomb worm Sabellariaalveolatareefson rockysubstrata andnoton soft sedi- Sabellariaalveolata(Desroyetal.,2011).Thisreefissituatedinthe ment.Reefsdevelopingonsoftsedimentarefarlessfrequentalong lowerintertidalzone(i.e.betweenthe-2andthe-4misobaths A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 3 (Noernbergetal.,2010)),paralleltothecoastandtothedominant at each station, six samples separated by at least 5 m were tidalcurrentsandalsonearimportantbluemussel(Mytilusedulis) randomly taken at low tide. The first three samples were done cultures.In2014,themaximaldimensionsoftheSainte-Annereef using a 18.5 cm side corer (269 cm2) to a depth of 15 cm (core were2.5kminlengthfor1kminwidthandtheengineeredsedi- samples).Forengineeredsediments,thisdepthcorrespondstothe ment represented about 32 ha for about 128 ha of associated layerwhereSabellariaalveolataandmorethan90%ofallspecies sediment(unpublishedresults).Thearealocatedinthecentralpart live(Duboisetal.,2002).Theotherthreesamplesweredoneusing ofthebayandalongthesameisobathasthereefischaracterizedby a1m2quadratinordertoestimatetheoverdispersedmacrofauna, medium to muddysands (Bonnot-Courtois et al., 2009) and bya mainly composed of bivalves and gastropods (quadrat samples). speciespoor“Macomabalthicacommunity”(Duboisetal.,2002). Allengineeredsedimentsamples(coreandquadratsamples)were taken at least 1 m from the reef edge to avoid a known border 2.2. Samplingdesignandlaboratoryanalyses effecton themacrofaunadiversity(Gruet,1972), whilethe asso- ciatedsedimentsamples(coreandquadratsamples)weretakenat Twosamplingareasweredefined;theSainte-Annereefareaand least 1 m away from the reef structures. The soft sediment core acontrolarea.Thereefareawascomposedoftwosedimenttypes, samplesweresievedthrougha1-mmsquaremeshonsitewhile theengineeredandtheassociatedsediments(Fig.1).Thecontrol the engineered sediment core samples were taken back to the areawasasoftsedimentzonelocated1.5kmNorth-Eastofthereef laboratory where they were broken apart under water and the areaandonthesamebathymetriclevel.Itwascharacteristicofthe faunaretainedona1-mmsquaremeshwascollected.Associated medium to muddy sands found in this part of the bay (Bonnot- andcontrolquadratsamplesweredonebysievingonsitethefirst Courtoisetal.,2009).Samplingtookplaceoveratwo-dayperiod 5 cm of sediment through a 5-mm square mesh. For the engi- in late winter (late February) and late summer (late September). neered quadrat samples, we sampled by hand all the visible Thesetwoseasonswerechosenbecausetheyarehighlycontrasted macrofaunalocatedonthereefandinsidethereefinterstices.All environmentally(e.g.hydro-sedimentaryfeatures)andbiologically coreandquadratsampleswerefixedina5%formaldehydesolu- (e.g.recruitmentpatterns,speciesturnover,growthrates).Indeed, tion, after which all the macrofauna was sorted, counted and winterisaperiodoflowbiologicalactivityandhighenvironmental identified to the species or genus level (except for nemerteans, pressures(coldtemperatures,windandstorms)whilelatesummer oligochaetes and nematodes) and finally preserved in a 70% is a post-recruitment period with a higher biological activity ethanol solution. For each engineered sediment core sample, all (ArbachLeloupetal.,2008;Cugieretal.,2010).Hence,samplingat theSabellariaalveolatawereweighted(totalwetweight). thesetwoseasonshelpsustohaveamorecompletepictureofthe To look at how the ecosystem engineer modifies its environ- dynamicshappeninginourdifferentstudyzones. ment,werandomlycollectedthreesedimentsamplesforgrain-size ToinvestigatetheeffectsofS.alveolataondiversityandspecies distribution,totalamountoforganicmatter(TOM),pigmentcon- composition, we compared the macrofauna associated with the centration (i.e. chlorophyll a and pheopigments) and total carbo- three different sediment types: the S. alveolata reefs, the sedi- hydrateconcentration(i.e.solubleandinsolublecarbohydrates),at mentspresentaroundthesestructuresandthecontrolsoftsedi- each associated and control sediment station. For the grain-size ments. For each sediment type (i.e. engineered, associated and distribution, the first 5 cm of sediments were sampled using a control sediment, Fig.1), ten stations were sampled. Everyengi- smallplastic core (19cm2). Foralltheothersedimentarycharac- neeredsedimentstationwaspairedwithanassociatedsediment teristics,onlythefirstcentimeterofsedimentwassampledusinga station,inordertoinvestigatehowthereefstructuresmodifythe plastic petri dish (57 cm2). Additional samples were collected in adjoiningsoftsediment.Thestationswereatleast75mapartand order to characterize the sediments constituting the Sabellaria Fig.1. Schematicoverviewpresentingthehabitatmodificationscausedby(1)theestablishmentofanecosystemengineerand(2)disturbancesoftheengineeredsediment. RecruitmentofS.alveolataleadstotheformationofabiologicallymodifiedsediment(engineeredsediment)andtoasoftsedimentundertheinfluenceoftheengineeredsediment (associatedsediment).Engineeredsedimentthenfacedirect(e.g.trampling,storms)and/orindirectdisturbances(e.g.shellfishfarming)whichcanleadtoagradualalteration. 4 A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 alveolatatubesaswellasthesedimentspotentiallytrappedwithin groupingofengineeredsedimentsamples.Speciespresentinonly thebiogenicstructure.Theseconsistedinrandomlycollectingthree onesample(i.e.inlessthan2%ofallsamples)wereexcludedfrom small reef parts (about 8 (cid:3) 3 cm) in each engineered sediment the initial matrix. To identify species typifying each sediment, station. Sediment grain-size distribution was obtained by me- speciesthatcorrelatedmorethan60%withoneofthefirsttwoaxes chanical sieving using AFNOR calibrated sieves (from 25 mm to (i.e.Spearmancorrelations)wereplottedoneachPCO.Inparallel,a 63 mm) and granulometric parameters were estimated using the one-way univariate permutational ANOVA (permanova) was per- ‘G2Sd’ package in R v. 3.3.0 (Fournier et al., 2014). Prior to me- formed on the same species density matrices as for the PCOs, in chanicalsieving,theengineeredsedimentswerecautiouslybroken ordertoevaluateiftherewasasignificantdifferenceinthespecies intotheiroriginalelements,i.e.mostlybioclastsasevidencedinLe compositionofeachsedimenttype. Cametal.(2011).Foralltheotheranalyses,thesedimentswerefirst Finally, the macrofauna diversity of each replicate (core and freeze-driedinordertoworkondrymatter.TOMwasdetermined associatedquadrat)sampledinlatewinterandlatesummer,was asthedifferencebetweentheweightoffreeze-driedsedimentand assessedusingHill'sindices;N0(numberofspecies),N1(exp(H0) theweightafter4hat450(cid:1) (AminotandKerouel,2004).Pigment where H0 is the Shannon-Winner diversity (loge)) and N2 (1/D concentrations (mg.g(cid:4)1 dry sediment) were estimated using the whereDistheSimpson'sdominanceindex(Hill,1973))asrecom- monochromatic technique (Lorenzen, 1967) described in Aminot mended by Gray (2000) and the total macrofauna density. These and Kerouel (2004). The chlorophyll a (Chl a) concentration was indicesinformhowthetotalabundanceispartitionedbetweenthe used as a proxy for microphytobenthos (MPB) biomass (Jeffrey differentspecies(Gray,2000;Whittaker,1972fordetails).Densities etal.,1997)whilepheopigments(Pheo)concentrationgaveusin- calculated using the CPUE method and for 1 m2 as previously formationabouttheamountofdegradedphotoautotrophs.Soluble detailed,wereusedtocalculateN1andN2.Foreachreplicate,N0 carbohydrates (Sol) present in the sediment were extracted by wascalculatedasthesumofthespeciesrichnessrecordedinthe hydrolysis(100(cid:1)Cfor45min),afterwhichthepelletsweretreated core andthespecies richnessrecorded inthe associated quadrat. with sulfuric acid (H2SO4) and placed 4 h at 100 (cid:1)C in order to ForN0,N1andN2,S.alveolatawaseitherkeptorremovedfromthe obtaintheinsolublecarbohydrates(Ins).SolandInsconcentrations initialdatainordertoinvestigatehowthisspeciesinfluencesthe (mg.g(cid:4)1drysediment)werethenestimatedbycolorimetricphenol partitioningoftheassociatedfaunaabundance. sulfuricdosage(Duboisetal.,1956).Solwereconsideredasbeing To test for significant differences between the three sediment an important labile source of carbon for consumers living in the typesforthedifferentgrainsizeandmacrofaunadescriptorsand sediment such as bacteria and deposit-feeding invertebrates becausenoneofthedescriptorsfulfillednormalityofdistribution (Bellingeretal.,2009)whiletheinsolublecarbohydratestosoluble and homogeneityof variance, permanovas were performed, with carbohydratesratio(Ins/Sol)wasusedasaproxyfortheC/Nratio sediment type considered as a fixed factor. We used Euclidian andasaTOMdegradationindex(Delmas,1983). distanceasadistancemeasureandran9999permutationsforeach test.Ifthemaintestwassignificant,pairwisetestswereperformed. 2.3. Dataanalysis Effectofthepresenceoftheengineeredsedimentonsoftsediment environmental parameters (TOM, Chl a, Pheo and Ins/Sol) was 2.3.1. Biologicalandenvironmentalengineeringeffects investigated by comparing these parameters between associated Sincemacrofaunawassampledusingtwodifferenttechniques andcontrolsediments,alsousingpermanovas.Priortoperforming (coresandquadrats),densitiesofspecieswereestimatedusingthe permanovas, we tested for homogeneity of dispersions using the catch-per-unit-effort (CPUE) method, i.e. the ratio between the PERMDISPPRIMERroutine(Andersonetal.,2008).Whenrawdata totalcatchandthetotalamountofeffortusedtoharvestthecatch presented significantly different dispersions between the three (Skalskietal.,2005).Atonesamplinglocation,whenaspecieswas sedimenttypes(p<0.05),itwaslogtransformed(inlatewinter: onlycollectedbycoreorquadrat,itsdensitywasestimatedusing principal mode, TOM, Chl a, Pheo, macrofauna density with and thecorrespondingsamplingsurface.However,whenaspecieswas withoutS.alveolata,N0withandwithoutS.alveolataandN2with sampledbybothmethods,cumulatedabundancesweredividedby S.alveolata,inlatesummer:macrofaunadensitywithandwithout thesumofeachgear'sCPUE.Thisestimationmethodwasusedfor S. alveolata, N0 with and without S. alveolata and N1 without 17speciesinlatewinterand15inlatesummer,takingintoaccount S.alveolata).Whenlogtransformationdidnotleadtohomogenous allthreesediment types.Species'densitieswerecalculated using dispersions(inlatewinter:%mud,%sandandSol,inlatesummer: theformula: TOM, Chl a, Sol, N1 and N2 calculated with S. alveolata), non- parametric statistical tests were performed (Kruskal-Wallis test (cid:1) (cid:3) ðabundanceAqþabundanceAcÞ forthegranulometricandmacrofaunaparametersandWilcoxon- densityA ind:m2 ¼ ðCPUEqþCPUEcÞ Mann-Whitneyfortheotherenvironmentalparameters). Inordertoevaluateifthedifferentenvironmentalandmacro- where densityA is species' A abundance per m2, abundanceAq is faunaparametersweresignificantlydifferentbetweenlatewinter species'Aabundanceusingthequadrat,abundanceAcisspecies'A and latesummer for each sediment type, one-factor permanovas abundance using the core, CPUEq is the quadrat's catch-per-unit- wereperformed,withseasonconsideredasafixedfactor.Wechose effort (1 m2) and CPUEc is the core's catch-per-unit-effort to perform one-factor rather than two-factor univariate analysis (0.0269m2). of variance (in this case with sediment type and season as fixed To assess the effect of Sabellaria alveolata on the associated factors), because we lacked replication inside each season for macrofauna and validate our a priori grouping into engineered, our different sediment types (Underwood, 1997). As previously associated and control sediments, Principal Coordinates Analysis mentioned, permanovas (9999 permutations) were used rather (PCO) were performed for the late winter and late summer data thant-testsbecausenoneof theinvestigatedvariableswerenor- sets. Analyses were performed on a Bray-Curtis similarity matrix mally distributed. Homogeneity of dispersions was also tested calculated from log-transformed densities after S. alveolata was (PERMDISP) and data was transformed when necessary (square- removed from the matrix, in order to take into account only the root transformation for TOM in the associated sediments, log speciesassociatedwiththissedimenttype.Indeed,becauseofits transformation for macrofauna density with S. alveolata in the highabundance(i.e.onaverage,63% of thetotalabundance),the controlsedimentsandformacrofaunadensitywithoutS.alveolata singlepresenceofS.alveolatawouldautomaticallycauseastrong intheengineeredsediments).ThePermanovas,PERMDISProutines A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 5 andPCOswereperformedusingthePRIMERv6softwarewiththe among regions (Bennett and Gilbert, 2016). We chose to use the PERMANOVAþadd-on(Andersonetal.,2008).Post-hocKruskal- presence/absence based indices presented by Baselga (2010) in Wallis tests were performed with the ‘kruskalmc’ function from order to partition total beta diversity, expressed by Sørensen the‘pgirmess’package(Giraudoux,2016)usingRversion3.3.0(R dissimilarity (bsor), into the turnover (bsim) and nestedness (bnes) CoreTeam,2016). components. In this case, bsor ¼ bsim þ bnes. Under conditions of equalspeciesrichness,bsor¼bsimandbnes¼0,whileundercon- 2.3.2. Linkingenvironmentalandbiologicalengineeringeffects ditions of unequal species richness, bsim and bnes vary between Therelationshipbetweentheenvironmentalcharacteristicsand 0 and bsor. Sørensen dissimilarity varies between 0 and 1, with the macrofauna present in the three sediment types wasinvesti- 0 indicating that two samples have identical species list and 1 gated using distance-based linear models (DistLM). In line with indicatingnocommonspecies(Baselga,2010).Forbsim,0indicates LegendreandAnderson(1999)andMcArdleandAnderson(2001), completenestedness,andamaximalvalueof1canbefoundifin DistLM models were coupled to a distance-based redundancy oneofthetwoconsideredsamples,therearenospeciesrecorded analysis (dbRDA) to define the best fitted model in a multi- and in the other, the number of species is maximal (Koleff et al., dimensional space in awaysimilar toa constrained PCO.DistLM 2003). To have a complementary vision of how disturbance modelswerebuiltusingtheBayesianInformationCriterion(BIC)to affected the associated fauna abundance, the abundance-based identify“good”modelsandthe‘best’proceduretoselectthevari- dissimilarity (Bray-Curtis dissimilarity, dBC) was also partitioned ablesaccordingtotheBIC.PriortotheDistLManddbRDAanalysis, intobalancedchangesinabundance(dBC-bal)andabundancegra- the environmental parameters were displayed using Draftsman dients(dBC-gra),whicharecloselyrelatedtoturnoverandnested- plotsandtheonespresentinganimportantskewnessweretrans- nesscomponentsrespectively(Baselga,2013).Theseindiceswere formedtoapproachnormality(Andersonetal.,2008).Iftwopre- computed after removing S. alveolata from the presence/absence dictorvariableswerestronglycorrelated(r2>0.80),oneof them and density matrices. They were calculated using the pairwise wasremovedfromtheanalysisinordertoavoidmulti-collinearity measuresinordertohavethebetadiversityandthedissimilarities (Dormann et al., 2013). Except for the grain-size data, environ- foreachpairofsamples(i.e.435pairs).Then,usingEuclidiandis- mental parameters used to characterize an engineered sediment tance, all the mud content pairwise differences were calculated. samplewerethesameasforitscorrespondingassociatedsediment Finally, using the different pairwise measures, we performed sample.Forlatewinter,thefinalpredictordatasetcontainedthe% Manteltests(9999permutations)forlatewinterandlatesummer sand, Pheo (both square-root transformed), % mud, TOM, data,totestthenullhypothesisofnorelationshipbetweenthemud S. alveolata biomass (all three fourth-root transformed), principal contentdistancematrixandeachbetadiversitymatrix.Ap-value modeandIns/Sol(bothlogtransformed).Forlatesummer,thefinal below 0.05 indicates a significant correlation between the two predictordatasetwasthesameasforlatewinter,exceptthe%sand investigated distance matrices, with the sign of the r-value indi- which was removed (absolute correlation with % mud > 0.8). cating if the twomatrices are positivelyor negativelyassociated. S.alveolatabiomasswasusedratherthanabundancebecausethis The beta diversity indices were computed using the ‘beta.pair’ parameterprovidesmoreinformationaboutecosystemfunctioning function, and the Bray-Curtis dissimilarity indices using the (Cardinale et al., 2013). S. alveolata biomass was considered as a ‘bray.part’ function, both from the ‘betapart’ R package (Baselga, predictorvariablesinceitphysicallymodifiesitsenvironmentand 2013). The Mantel tests were performed using the ‘mantel.rtest’ it was consequently removed from the macrofauna data set. The functionfromthe‘ade4’Rpackage(DrayandDufour,2007). DistLM models and dbRDA analysis were performed using the Totestthelinkbetweenthemacrofaunalassemblagesbasedon PRIMER v6 software with the PERMANOVA þ add-on (Anderson their respective beta diversity and dissimilarity indices and the etal.,2008). disturbance parameter (i.e. mud content), non-metric multidi- mensional scaling ordinations (nMDS) were successively per- 2.3.3. Disturbancesandbiologicalengineeringeffect formedforeachindex(bsor,bsim,bnes,dBC,dBC-balanddBC-gra)andat Atitsclimax,aS.alveolatareefisformedby100%honeycomb each sampling period (late winter and late summer) using the worm tubes, leaving virtually no space for infaunal organisms. ‘metaMDS’functionofthe‘MASS’Rpackage(VenablesandRipley, When natural or anthropogenic disturbances (e.g. storms, tram- 2002).Then,the‘envfit’function(‘vegan’Rpackage)wasusedto pling)physicallydamagethereef,tubesaredestroyed,freeingup test if the mud content was significantly correlated with each space.Thisnewavailablespacecanbefilledeitherwithotheror- ordination(Oksanenetal.,2016).Whenacorrelationwassignifi- ganisms such as the oyster Magallana gigas (formerly known as cant,themudcontentswerefittedandplottedonthegivennMDS Crassostrea gigas) or by fine particles. Fine particles accumulate using the ‘ordisurf’ function of the ‘vegan’ R package (Oksanen fromsuspendedsediments, orfromthefecesandpseudofecesof et al., 2016). All these analyses were performed using R version S. alveolata and other bivalves (biodeposition) (Dubois et al., 3.3.0(RCoreTeam,2016). 2006b). In eithercase, this finesediment can end uptrapped in- sidetheS.alveolatareefs.Consequently,theincreaseddepositionof 3. Results mud inside the engineered sediments is the result of several different and often concomitant disturbances. Fine sediment 3.1. Environmentalengineeringeffect depositionhaspreviouslybeenrecognizedasasignificantdistur- bance to stream macroinvertebrates (Mathers et al., 2017) and Mean values of grain-size distribution parameters measured benthic habitats (Balata et al., 2007; Mateos-Molina et al., 2015; within each sediment type are reported in Table 1a. Analyses Milleretal.,2002).Similarly,wechosetoconsidermudcontentasa revealedsignificantdifferencesbetweenthesedimenttypesforall proxy for disturbance. This proxy was also chosen because it is testedmetricsinlatewinter(p<0.001)andforallbutoneinlate independent from Sabellaria alveolata population dynamics and summer(mudcontent).Atbothperiods,therewasastrongengi- physiologicalstate.Finally,usingthemudcontentmakesthetwo neering effect on the principal mode marked by a significantly seasonsreadilycomparable. coarsersedimentintheengineeredandassociatedsedimentsthan Beta diversity was calculated using pairwise multivariate dis- inthecontrolsediments(Table1a).Inlatewinter,thesortingindex tancessincetheyareindependentofsamplesizeandregionaldi- S0 was significantlylowerin the engineered and associated sedi- versity(gammadiversity)allowingaccuratepotentialcomparisons mentsthaninthecontrolandmudcontentwassignificantlylower 6 A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 Table1 Meanvalues(±standarderrors)for(a)thegrain-sizeparametersofthethreesedimenttypes(engineered,associatedandcontrol)and(b)theenvironmentalparametersforthe associatedandthecontrolsediments.Significantdifferences(p<0.05)oftheone-wayANOVAsareinboldandfor(a),post-hocresultsaredesignatedbysuperscriptletters indicatinghomogenousgroupsofsamples.TOM:totalorganicmattercontent,Chla:chlorophyllaconcentration,Pheo:pheopigmentsconcentration,Sol:solublecarbo- hydratesconcentration,Ins/Sol:ratiooftheconcentrationofinsolublecarbohydratesonsolublecarbohydrates. (a) Latewinter Latesummer Engineered Associated Control p-value Engineered Associated Control p-value Principalmode(mm) 688±35a 1010±118a 186±8b <0.001 618±8a 692±74a 201±9b <0.001 Sortingindex(S0) 1.71±0.05a 1.72±0.05a 2.97±0.34b <0.001 1.69±0.05a 2.98±0.45b 2.70±0.37b 0.018 Mud(%)(<63mm) 10.00±0.83a 1.84±0.44b 27.38±3.62a <0.001 9.59±1.22a 20.47±5.37a 21.61±5.23a 0.106 Sand(%)(63e200mm) 87.19±0.83a 76.74±1.40b 71.69±3.53b <0.001 85.77±1.40a 65.11±4.09b 76.79±5.17ab 0.001 (b) Latewinter Latesummer Associated Control p-value Associated Control p-value TOM(%) 6.96±0.72 2.70±0.30 <0.001 4.91±0.59 2.26±0.28 <0.001 Chla(mg.g(cid:4)1sediment) 12.21±2.49 2.83±0.58 0.0022 13.39±2.24 3.92±0.88 0.002 Pheo(mg.g(cid:4)1sediment) 14.54±0.36 16.18±0.36 0.0014 15.56±0.53 15.41±0.29 0.826 Sol(mg.g(cid:4)1sediment) 442±72 113±25 0.0027 467±78 120±25 <0.001 Ins/Sol 8.59±2.29 8.63±0.37 0.9998 5.96±0.43 6.32±0.33 0.5175 intheassociatedsedimentsthanintheothertwosedimenttypes. (permanova: p ¼ 0.28) probably because of the important vari- Finally,thesandcontentwassignificantlyhigherintheengineered abilityinlatewinter(Table1). sediment relative to the other sediment types. In late summer, associated sediments had a higher sorting index than the engi- 3.2. Biologicalengineeringeffect neered sediments and one comparable to the control sediments. Althoughassociatedsedimentswerealsocharacterizedbyahigher In late winter, 9244 organisms belonging to 121 different taxa mudcontentinlatesummercomparedtolatewinter(permanova: were sampled in the cores and 8478 organisms belonging to 26 p ¼ 0.0051), no significant difference was observed between the differenttaxaweresampledwiththequadrats(seetheAppendixfor three sediment types. For all grain-size parameters, the control acompletelistofspecies).Comparatively,inlatesummermoreor- sedimentsshowednosignificantchangesbetweenlatewinterand ganismsandtaxaweresampledwiththecores(23463organisms/ latesummer(permanova:p(principalmode)¼0.23,p(S0)¼0.60, 125taxa)whilefewerorganismsandmoretaxaweresampledwith p(mud)¼0.37andp(sand)¼0.42).Thepatternwassimilarforthe thequadrats(4677organisms/30taxa).Forallsedimenttypes,total engineered sediments (permanova: p(principal mode) ¼ 0.059, species richness was consistently higher in late summer than in p(S0) ¼ 0.78, p(mud) ¼ 0.78 and p(sand) ¼ 0.39). The associated late winter but this differencewas significantonly for the control sedimentsshowedsignificantchangesintheirgrain-sizedistribu- and engineered sediments (permanova: p(control) ¼ 0.039, tionbetweenlatewinterandlatesummer.Inlatewinter,theywere p(associated)¼0.071andp(engineered)¼0.0001). muchmorehomogenousthaninlatesummer(Table1)andthey PCOsandone-waypermanovasperformedondensitymatrices became significantly muddier between the two sampling cam- indicated that the three sediment types significantly differed paigns(permanova:p¼0.0051)leadingtoasignificantdecreasein (p < 0.05) in their associated fauna at both sampling periods, theprincipalmode(permanova¼0.025). confirmingourapriorisedimenttypegrouping(Fig.2andFig.3). The comparison of sedimentary parameters revealed a strong PCOaxis1explainedinlatewinterandlatesummer,respectively engineering effect at both periods regarding TOM, Chl a and Sol 26.1 and 30.3% of the total variation present in the resemblance (Table1b,p<0.005).Inbothseasons,TOMwasconsistentlytwice matrix and clearly separated the engineered samples from the as high in the engineered environment than in the control zone. control samples. PCO axis 2 explained in late winter and late Organicmattercontentalsoshowedasignificantdecreasebetween summer, respectively 14.6 and 14.8% of the total variation and late winter and late summer in the reef zone (permanova: discriminatedtheengineeredandcontrolsamplesfromtheasso- p¼0.029)andnosignificanttemporalchangeinthecontrolsed- ciated samples. In both seasons, engineered samples werehighly iments(permanova:p¼0.29).Similarly,Chlaconcentrationwas clusteredcompared tothe more scattered associated and control tentimeshigherinthesoftsedimentsadjacenttotheengineered sedimentssamples.Inlatewinter,thecontrolandassociatedsed- structures than in the control and did not displayany significant iments were well separated while there was a small overlap be- temporalchanges in either the control (permanova: p ¼ 0.29) or tween the associated and engineered sediments (Fig. 2). In late theassociatedsediments(permanova:p¼0.72).Solconcentration summer,therewasanoverlapbetweentheassociatedandcontrol was also consistently four times higher in the reef environment sediments (Fig. 3). This overlap was mostly due to bivalves like thaninthecontrolanddisplayedatemporalstabilitysimilartothe Limecola balthica or Cerastoderma edule and to the polychaete Chla(permanova:p(control)¼0.87andp(associated)¼0.82).In Nephtyshombergii(Fig.3andAppendix).Finally,engineeredsedi- latewinter,thePheoconcentrationwassignificantlyhigherinthe ments were characterized by a much greater number of species control than in the associated sediments while in late summer, correlatedatmorethan60%witheachPCOaxis(11inlatewinter there was no significant difference. In both sediment types, and17inlatesummer)thantheassociated(3inlatewinterand1in Pheo concentrations did not show significant changes between latesummer)andthecontrolsediments(3inlatewinterand6in the two sampling campaigns (permanova: p(control) ¼ 0.10 and latesummer). p(associated)¼0.11).Finally,Ins/Solwasnotsignificantlydifferent Mean macrofauna diversity indices and densities were calcu- betweenassociatedandcontrolsedimentsinlatewinterandlate latedwithineachsedimenttypeandforeachsamplingcampaign summer, andwas significantly higher in latewintercompared to (Table2aandb).Atthesedimenttypescale,one-waypermanovas latesummerforthecontrolsediments(permanova:p¼0.0001). showedsignificantdifferencesbetweenengineeredsedimentson Thistemporalpatternwasnotdetectedintheassociatedsediments theonehandandassociatedandcontrolsedimentsontheother,for A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 7 allthediversitymeasurementsanddensitiesatbothperiods.There weretwoexceptionsregardingN1andN2calculatedinlatesum- merwithS.alveolatatakenintoaccount.Inthesecases,therewere nosignificantdifferencesbetweenthethreesedimenttypes.When S.alveolatawastakenintoaccount,totalmacrofaunadensitywas 20timeshigherintheengineeredsedimentsatbothperiods.This differencewasmaintainedevenafterS.alveolatawasremovedfrom thedatasetbutitwasreducedtoanaverage5-folddifference.The engineeredsedimentwasalsohometosignificantlymorespecies (meanspeciesrichnessN0)thantheassociated andcontrol sedi- mentsandthis,whateverthesituation. Regarding macrofauna density, N1 and N2, associated and control sediments presented similar temporal patterns when comparing late winter and late summer. Their respective macro- faunadensityincreasedsignificantlybetweenthetwocampaigns (permanova:p(control)¼0.023andp(associated)¼0.018)while N1 and N2 showed non-significant differences (permanova: p(control-N1) ¼ 0.15, p(control-N2) ¼ 0.25, p(associated- N1)¼0.83andp(associated-N2)¼0.53).Betweenlatewinterand late summer, the engineered sediments presented a significant increase in the total macrofauna density (permanova: p(density Fig.2. PCOanalysisofmacrobenthosassociatedwiththethreesedimenttypesinlate withS.alveolata)¼0.0001)onlydrivenbyasignificantincreasein winter.TheanalysisisbasedonBray-Curtissimilaritiesoflogtransformeddensity the associated fauna density (permanova: p(density without data.Theblackdiamonds,thegreysquaresandthelightgreycirclesrepresentthe S.alveolata)¼0.0001andp(S.alveolatadensity)¼0.54).Theyalso engineered, the associated and the control sediment samples respectively. Vectors representspeciescorrelatingmorethan60%withoneofthefirsttwoPCOaxes.The showedasignificantincreaseinthecaseofN1andN2calculated correlationsarebasedonSpearmancoefficients.ASIM:Acheliasimplex,CEDU:Cera- withS.alveolata(permanova:p(N1)¼0.0007andp(N2)¼0.0001), stodermaedule,CFOR:Crepidulafornicata,CMAE:Carcinusmaenas,CVOL:Corophium achangewhichwasnotsignificantoncetheengineerspecieswas volutator,GBOB:Goniadellabobrezkii,GUMB:Gibbulaumbilicalis,GVUL:Golfingiavul- removed(permanova:p(N1)¼0.089andp(N2)¼0.73). garis, LBAL: Limecola balthica, LLEV: Lekanesphaera levii, LRUG: Lekanesphaera rugi- cauda, McfGAL: Mytilus cf. galloprovincialis, MFRA: Mediomastus fragilis, MGIG: Magallanagigas,MPAL:Melitapalmata,NCIR:Nephtyscirrosa,NLAP:Nucellalapillus, 3.3. Linkingenvironmentalandbiologicalengineeringeffects PCUL:Perinereiscultrifera,PPLA:Porcellanaplatycheles. DistLM and dbRDA analysis were performed in late winter (Fig. 4a) and late summer (Fig. 4b) with S. alveolata biomass considered as an environmental parameter. In both seasons, S.alveolatabiomasswastheparameter whichbestexplainedthe relationship between environmental parameters and macrofauna assemblages(18.0%inlatewinterand24.8%inlatesummer).Inlate winter,themostparsimoniousmodel,explaining33.6%ofthetotal variation in species assemblages, was defined by (1) Sabellaria biomass(square-roottransformed,18.0%),(2)principalmode(log transformed,13.2%) and (3) total organic matter content (fourth- roottransformed,10.7%,Fig.4).Thefirsttwoaxesexplained91.6% of the fitted variation and 30.7% of the total variation. Species assemblage werestructured according totwo gradients. The first was driven by S. alveolata, and separated engineered sediments fromthetwoothertypes.Thesecondwasdrivenbythesediment principalmodeandthetotalorganicmattercontentandseparated theassociatedfromthecontrolsediments(Fig.4a).Inlatesummer, themostparsimoniousmodelexplained40.7%ofthetotalvariation inspeciesassemblages.Itwasdefinedbythesamefirsttwovari- ables as for late winter: Sabellaria biomass (square-root trans- formed, 24.8%) and principal mode (log transformed,16.9%). The third selected variable differed from late winter since it was the Fig.3. PCOanalysisofmacrobenthosassociatedwiththethreesedimenttypesinlate mudcontent(fourth-roottransformed)anditexplainedonlyavery summer.TheanalysisisbasedonBray-Curtissimilaritiesoflogtransformeddensity small part of the total variation (0.079%). The first two axes data.Theblackdiamonds,thegreysquaresandthelightgreycirclesrepresentthe engineered, the associated and the control sediment samples respectively. Vectors explained87.5%ofthefittedvariationand35.6%ofthetotalvari- representspeciescorrelatingmorethan60%withoneofthefirsttwoPCOaxes.The ation.Again,speciesassemblageswerestructuredaccordingtotwo correlationsarebasedonSpearmancoefficients.AECH:Acheliaechinata,ALAE:Achelia gradientsbuttheydidnotseparatethedifferentsedimenttypesas laevis,ASIM:Acheliasimplex,CEDU:Cerastodermaedule,CMAE:Carcinusmaenas,CVOL: clearlyasinlatewinter.S.alveolatastilldefinedthefirstgradient Corophiumvolutator,EORN:Eulaliaornata,GBOB:Goniadellabobrezkii,GMAX:Gnathia maxillaris,GUMB:Gibbulaumbilicalis,GVUL:Golfingiavulgaris,LBAL:Limecolabalthica, andclearlyseparatedtheengineeredsedimentsfromthetwosoft LCON: Lanice conchilega, LLEV: Lekanesphaeralevii, LRUG: Lekanesphaera rugicauda, sediments. The opposition between the principal mode and the MARE: Malmgrenia arenicolae, McfGAL: Mytilus cf. galloprovincialis, MFRA: Medi- mudcontentdefinedthesecondgradient.Alongthisgradient,the omastusfragilis,MGIG:Magallanagigas,MPAL:Melitapalmata,NCIR:Nephtyscirrosa, distinction associated/control sediments was not well defined. NEMA:Nematodaspp.,NEME:Nemertesp.,NHOM:Nephtyshombergii,NLAP:Nucella Indeed, there were three associated sediment samples character- lapillus,NMIN:Nephasomaminutum,OCTE:Odontosyllisctenostoma,PCUL:Perinereis cultrifera,PPLA:Porcellanaplatycheles. ized by high mud contents and isolated from the rest of the 8 A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 Table2 Meanvalues(±standarderrors)forthetotalmacrofaunadensity(numberofindividuals.m(cid:4)2),N0,N1andN2with(a)Sabellariatakenintoaccountand(b)Sabellariaexcluded, forthethreesedimenttypes(engineered,associatedandcontrol)andatbothsamplingperiods(latewinterandlatesummer).N0representsthespeciesrichness,N1the exponentialoftheShannon-WinnerdiversityandN2theinverseoftheSimpsondominanceindex.Significantdifferences(p<0.05)oftheone-wayANOVAsareinboldand post-hocresultsaredesignatedbysuperscriptlettersindicatinghomogenousgroupsofsamples. Latewinter Latesummer (a)Macrofauna(Sabellariaincludedintheanalyses) Engineered Associated Control p-value Engineered Associated Control p-value Density 10067±841a 585±102b 629±109b <0.001 23911±2530a 1029±156b 1403±351b <0.001 N0 17±1a 7±1b 8±1b <0.001 26±1a 9±1b 10±1b <0.001 N1 2.92±0.37a 4.46±0.50b 4.54±0.37b 0.013 6.01±0.65a 4.61±0.38a 5.22±0.28a 0.229 N2 1.87±0.23a 3.75±0.40b 3.60±0.28b <0.001 3.93±0.44a 3.44±0.30a 4.04±0.25a 0.315 (b)Macrofauna(Sabellariaexcludedfromtheanalyses) Engineered Associated Control p-value Engineered Associated Control p-value Density 2385±518a 538±91b 629±109b <0.001 11066±1814a 981±137b 1403±351b <0.001 N0 16±1a 7±1b 8±1b <0.001 25±1a 9±1b 10±1b <0.001 N1 7.73±0.51a 4.30±0.49b 4.54±0.37b <0.001 9.00±0.52a 4.51±0.37b 5.22±0.28b <0.001 N2 5.63±0.42a 3.64±0.39b 3.60±0.28b <0.001 5.82±0.38a 3.36±0.30b 4.04±0.25b <0.001 associatedsedimentsamples(Fig.4b). (r2¼41.33%),andbetween20and30%ofbnes(r2¼21.25%)anddBC- bal (r2 ¼ 29.56%). When the correlationwas significant, the fitted mud contents were plotted on the corresponding nMDS plots 3.4. Disturbancesandbiologicalengineeringeffect (Figs.5and6).Thecorrelationbetweenthedisturbanceproxyand the different nMDS plots showed a pattern similar to the one Consistent mean values in late winter (10%) and late summer (9.59%), confirm the choice of the mud content as a suitable revealed by the late summer Mantel test, with beta diversity ‘disturbance parameter’ (Table 1a). Indeed, these values did not changes mainly driven by a species turnover and an abundance significantlyvarybetweenthetwocontrastedseasonswesampled gradient. (permanova: p ¼ 0.78). Incontrast,themean S. alveolatadensity almost doubled between late winter (7682 ± 3312 ind.m(cid:4)2) and 4. Discussion latesummer(12844±14262ind.m(cid:4)2),withaveryhighsummer variability,leadingtonosignificantchange(permanova:p¼0.54). 4.1. Engineeredstructurescausegrain-sizedistributionchanges Oppositely, the mean S. alveolata biomass bysurface unit signifi- cantlydecreasedbetweenlatewinter(646±317gm(cid:4)2)andlate Environmentalengineeringeffectsarecomposedoftwotypesof summer(318±211gm(cid:4)2)(permanova:p¼0.0001). changes, structural and abiotic changes, structural changes being Mantel tests performed between the mud content distance causedbyecosystemengineersandinducingabioticchanges(Jones matrix and the different beta diversity matrices showed a clear et al., 2010). S. alveolata is capable of biologically modifying soft temporaldifferencebetweenlatewinterandlatesummer.Thetests sedimentsbyselectivelygluingtogetherbioclasticsandparticles,in werenotsignificantwhenperformedusingthelatewinterdatasets order to build its tube (Fournier et al., 2010). This leads to the (p> 0.05,Table3),whiletheyrevealedasignificantandpositive transformationofaninitialsoftsedimentintoathree-dimensional correlation between the mud content distance matrix and bsor hardsubstratumwithalonglastingresistancetophysicalloading (p < 0.001, r ¼ 0.24), bsim (p ¼ 0.0066, r ¼ 0.15), dBC (p < 0.001, via the secreted organic cement (Le Cam et al., 2011). Sabellaria r¼0.38)anddBC-gra(p<0.001,r¼0.29)(Table3)usingthelate alveolata can therefore be considered as a “structural engineer” summer data sets. These results indicate that in late winter, an according to Berke (2010). Structural changes caused by physical increaseinmudcontent,usedasaproxyfordisturbance,doesnot ecosystemengineersresultinavariationinthedistributionoffluid leadtobetadiversitychangesbutinlatesummer,itleadsto(1)an andsolidmaterialtermedabioticchanges(Jonesetal.,2010).Inthe increaseinbetadiversitydrivenbyaspeciesreplacementand(2) case of S. alveolata, a direct abiotic engineering effect observable an increase in abundance-based dissimilarity driven byan abun- throughtheengineeredsedimentsandanindirectone,observable dancegradient.Ordinationplotsofsimilarities(nMDS)ofmacro- throughtheassociatedsediments,weredetected.Engineeredand faunalassemblagesbasedonbsor,bsim,bnes,dBC,dBC-balanddBC-gra associated sediments presented, at both sampling periods, a indiceswereperformedinlatewinterandlatesummer(Figs.5and coarsertexturethanthecontrolsediments,confirmingtheimpact 6).Inlatewinter,thecorrelationbetweenthemudcontentandthe Sabellariidaepolychaetes haveonthelocalsediment's texture by different nMDS plots was significant for bsor (p ¼ 0.008), bnes selecting sand particles of a specific size to build their tubes (p¼0.023),dBC(p¼0.019)anddBC-gra(p¼0.027).Themudcontent (Phragmatopoma caudata (¼ P. lapidosa) (Gram,1968; Kirtleyand explained30.67%oftheordinationbasedonbsorand24.54%ofthe Tanner,1968; Main and Nelson,1988), Sabellaria vulgaris (Wells, ordinationbasedonbnes.Similarly,26.93%and24.51%oftheordi- 1970), Sabellaria nanella (Bremec et al., 2013)). Ultimately, these nationbasedondBCanddBC-grarespectivelywhereexplainedbythe bioconstructingSabellariidaespeciescreatereefscharacterizedby mud content. In late summer, the correlation between the mud agrain-sizedistributiondifferentfromthelocalsoftsediments.The content and the different nMDS plots was significant and much caseoftheassociatedsedimentsraisesthequestionofthedefini- higher for all the indices; bsor (p ¼ 0.001), bnes (p ¼ 0.036), bsim tionofareefhabitat.InEurope,“reefs”arerecognizedasamarine (p ¼ 0.001), dBC (p ¼ 0.001), dBC-gra (p ¼ 0.002) and dBC-bal habitat to be protected and are listed under Annex I of the EC (p ¼ 0.006). Indeed, the mud content explained over 50% of the Habitats Directive (Council Directive EEC/92/43 on the Conserva- ordinationbasedonbsor(r2¼53.07%)anddBC(r2¼52.76%),around tion of Natural Habitats and of Wild Fauna and Flora) under the 40% of the ordination based on bsim (r2 ¼ 39.23%) and dBC-gra designation of Special Areas of Conservation (SACs). They are A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 9 habitats presenting stable sedimentary conditions. The mud pre- sent in the engineered sediments is located in small cracks and crevices protected from the main hydrodynamic processes (i.e. winterstorms,tidalcurrentsandswell).Conversely,theassociated sedimentsarecharacterizedbyasteepandsignificantincreasein mudcontentbetweenwinter(2%)andsummer(21%).Asshownby Calineetal.(1988)fortheSainte-Annereef(MSMB),localizedmud depositions are linked to hydrodynamic and associated hydro- sedimentary processes induced by the presence of the reef itself andofthemusselfarms(bouchots)infrontofthereef(McKindsey et al., 2011). These mud depositions are observed behind reef structures important enough to act as physical barriers (Caline et al., 1988), where they are generally superficial and conse- quentlyeasilyerodedbystrongwaveaction,limitingtheirpresence inwinter. 4.2. Engineeredstructuresenhancebenthicprimaryproductionand potentiallymicrobialactivity As reported by Jones et al. (2010), abiotic changes induced by physicalengineeringactivitycanthemselvescausebioticchanges. Ourresultsclearlyshowthatatbothseasons,associatedsediments have a higher organic mattercontent compared with the control sediments. At both seasons, high levels of organic matter were associatedwithhighchlorophyllaconcentrations,indicatingthat partoftheorganicmatterpresentintheassociatedsedimentsisthe consequenceofMPBdevelopment.Thehighbenthicprimarypro- ductionpromotedbytheSainte-Annereef,comparedtoagenerally lower benthic production in the MSMB as measured by Davoult etal.(2008)andMigne(cid:1)etal.(2009),confirmsitsimportantbiotic engineering effect. Similar results were found for the invading intertidal reef-forming polychaete Ficopomatus enigmaticus (Bruschettietal.,2011),forshallowoysterreefs(Crassostreavirgina, Newell et al., 2002) and for intertidal mussel beds (Engel et al., 2017). According to Berke (2010), “structural engineers operate through similar processes and have similar types of effects”. Consequently,thecreationofbenthicprimaryproductionhotspots byreef-buildingstructuralengineerscouldbeageneralpropertyof thesemarinespecies.Nonetheless,thisphenomenonwasobserved at the scale of the largest and probablyoldest S. alveolata reefin Europe(AudouinandMilne-Edwards,1832)andthestudybyEngel etal.(2017)highlightedtheimportanceofthesizeandageofthe bioconstruction in promoting local benthic microalgae. Hence, furtherstudiesareneededtoconfirmthegeneralroleofS.alveolata reefsas“biologicalpowerstations”(Engeletal.,2017). Fig.4. dbRDAplotsbasedona)thelatewinterdatasetandb)thelatesummerdata Furthermore,thehighchlorophyllaconcentrationsmeasuredin setandrepresentingthethreesedimenttypemacrofaunacompositionasexplainedby thesetofenvironmentalparameterscomposingthemostparsimoniousexplanatory latewinterandlatesummerindicatethatS.alveolatareefspromote model. Vectors represent the environmental parameters selected by the DistLM animportantbenthicprimaryproductionallyearround,thatcould routine.Theblackdiamonds,thegreysquaresandthelightgreycirclesrepresentthe be a relevant food source for deposit- (Kanaya et al., 2008) and engineered,theassociatedandthecontrolsedimentsamplesrespectively. suspension-feeders (Lefebvre et al., 2009) through resuspension processes(Hylleberg,1975;Ubertinietal.,2015).Intheassociated definedas“submarineorexposedatlowtide,rockysubstratesand sediments,MPBoftengrowsonsmallaccumulationsofpuremud biogenicconcretions”.Inthelightofourfindings,wecanverywell andisconsequentlyeasilyerodedandavailabletoconsumers.Such considertheengineeredandtheassociatedsedimentsasthesame benthicprimaryproductionmayhaveatrophicimportanceduring sediment but under two different forms, a consolidated (engi- thewintermonths(Lefebvreetal.,2009),whenthephytoplankton neered sediments) and an unconsolidated form (associated sedi- productionistypicallylow(ArbachLeloupetal.,2008;Cugieretal., ments). Hence, we propose to widen the definition of a “reef” to 2010).FilterfeedingmollusksareknowntostimulateMPBgrowth includethenon-engineeredsedimentsunderitsdirectinfluence. (Engeletal.,2017;Newelletal.,2002)viainorganicnutrientrelease The main difference between the engineered and associated (i.e.carbon,nitrogenandphosphorus(vanBroekhovenetal.,2014)) sedimentsconcernstheirmudcontent.Atbothseasons,theengi- and bacterial remineralization of their biodeposits (van neered sediments havea mean mudcontent around 10%, as pre- Broekhoven et al., 2015). Similarly, S. alveolata produces large viouslyobservedbyLeCametal.(2011).Sabellariawilsoniveneers amountsoffecesandpseudofecesvisibleonthesediment(Dubois havealsobeenreportedtopresentconsistentsiltandclaycontents et al., 2005), that could favor MPB growth. Primary production across two contrasting seasons (rainy and dry seasons in Ataide could also be enhanced by the presence of other suspension- et al., 2014) indicating that Sabellariidae polychaetes build new feeders living in the engineered sediments, such as Magallana 10 A.G.Jonesetal./Estuarine,CoastalandShelfScience200(2018)1e18 Table3 ResultsoftheManteltestsbetween(a)thedifferentbetadiversitymatricesandthemudcontentdistancematrixand(b)thedifferentabundance-baseddissimilaritymatrices andthemudcontentdistancematrixatbothsamplingperiods(latewinterandlatesummer).bsoristheSørensenpairwisedissimilarityandaccountsforthetotalbetadi- versity,bsimistheSimpsonpairwisedissimilarityandaccountsfortheturnovercomponentofthetotalbetadiversity,bnesisthenestedness-resultantdissimilarityandac- countsforthenestednesscomponentofthetotalbetadiversity;bsor¼bsimþbnes.dBCistheBray-Curtisindexofdissimilarityandaccountsforthetotalabundance-based dissimilarity,dBC-balisthebalancedvariationinabundancescomponentoftheBray-Curtisdissimilarityandisequivalenttoanabundance-basedturnover,dBC-graisthe abundancegradientcomponentofBray-Curtisdissimilarityandisequivalenttoanabundance-basednestedness;dBC¼dBC-balþdBC-gra.Significantsimulatedp-values (p<0.05)andassociatedobservedcorrelationareinbold. Latewinter Latesummer Observedcorrelationr Simulatedp-value Observedcorrelationr Simulatedp-value (a)Betadiversityindices bsor 0.13 0.070 0.24 <0.001 bsim 0.066 0.23 0.15 0.0066 bnes 0.032 0.33 0.077 0.094 (b)Abundance-baseddissimilarityindices dBC 0.14 0.052 0.38 <0.001 dBC-bal 0.050 0.28 0.058 0.18 dBC-gra 0.046 0.28 0.29 <0.001 Fig.5. LatewinternMDSordinationplotsofthebenthicmacrofaunaassemblagesbasedona)theSørensentotalbetadiversity,b)thenestednesscomponentofthetotalbeta diversity,c)theBray-Curtisindexofdissimilarityandd)theabundancegradientcomponentoftheBray-Curtisdissimilarity.ThestressvalueofthenMDSisindicatedoneachplot. Thelinesindicatethedifferentfittedmudcontentsobtainedusingthe‘ordisurf’function. gigas,whichcanreachdensitiesof100ind.m(cid:4)2asmeasuredinthe these pseudofeces are deposited on the associated sediments, it disturbed engineered sediments using the quadrats. As already couldincreasetheirconcentrationinsolublecarbohydrates.Solu- observedinFicopomatusenigmaticusreefs(Bruschettietal.,2011), ble carbohydrates also compose the extracellular polymeric sub- S.alveolatareefsprobablyincreasethebentho-pelagiccouplingby stancesproducedbybenthicdiatoms(Bellingeretal.,2009)andare linkingpelagicorganicmattertothebenthiccompartmentviatheir animportantsourceoforganiccarbon,rapidlyconsumedbyhet- suspension-feedingactivityandbiodeposition. erotrophicmicroorganismspresentinthesediment(Bhaskarand In late winter and late summer, associated sediments had Bhosle, 2005; Goto et al., 2001). Consequently, S. alveolata pres- consistently higher soluble carbohydrate concentrations than the ence could support allyear round an important bacterial activity controlsediments.Carbohydratesarethecomponentsofthemucus through the soluble carbohydrates excreted by the diatoms and coating the pseudofeces produced by S. alveolata and other presentinthemucuscoatingthebiodeposits.Thisorganiccarbon suspension-feeders (van Broekhoven et al., 2015). Hence, when can either be used by the bacteria for their growth (bacterial
Description: