Wave stress and biotic facilitation drive community ...

https://helda.helsinki.fi

Wave stress and biotic facilitation drive community composition

in a marginal hard-bottom ecosystem

Westerbom, Mats

2019-10-01

Westerbom , M , Kraufvelin , P , Erlandsson , J , Korpinen , S , Mustonen , O & Díaz , E

2019 , ' Wave stress and biotic facilitation drive community composition in a marginal

hard-bottom ecosystem ' , Ecosphere , vol. 10 , no. 10 , 02883 , pp. e02883 . https://doi.org/10.1002/ecs2.2883

http://hdl.handle.net/10138/311276

https://doi.org/10.1002/ecs2.2883

cc_by

publishedVersion

Downloaded from Helda, University of Helsinki institutional repository.

This is an electronic reprint of the original article.

This reprint may differ from the original in pagination and typographic detail.

Please cite the original version.

Wave stress and biotic facilitation drive community composition ina marginal hard-bottom ecosystem

MATS WESTERBOM,1,� PATRIK KRAUFVELIN,2 JOHAN ERLANDSSON,3 SAMULI KORPINEN,4

OLLI MUSTONEN,1 AND ELIECER D�IAZ5

1Tv€arminne Zoological Station, Helsinki University, J. A. Palmensv€ag 260, Hang€o 10900 Finland2Department of Aquatic Resources, Institute of Coastal Research, Swedish University of Agricultural Sciences, Skolgatan 6, €Oregrund

74242 Sweden3Milj€of€orvaltningen, G€oteborgs Stad, Box 7012, G€oteborg 402 31 Sweden

4Finnish Environment Institute, Marine Research Centre, Latokartanonkaari 11, Helsinki 00790 Finland5Department of Environmental Sciences, University of Helsinki, PO Box 65 (Viikinkaari 1), Helsinki 00014 Finland

Citation: Westerbom, M., P. Kraufvelin, J. Erlandsson, S. Korpinen, O. Mustonen, and E. D�ıaz. 2019. Wave stress andbiotic facilitation drive community composition in a marginal hard-bottom ecosystem. Ecosphere 10(10):e02883. 10.1002/ecs2.2883

Abstract. Ecological patterns are inherently scale-dependent and driven by the interplay of abiotic gradi-ents and biotic processes. Despite the fundamental importance of such gradients, there are many gaps inour understanding of how abiotic stress gradients interplay with biotic processes and how these collec-tively affect species distributions. Using a hierarchical design, we sampled two communities separated bydepth along wave exposure and salinity gradients to elucidate how these two gradients affect species com-position in habitats formed by the foundation species Mytilus trossulus and Fucus vesiculosus. Specifically,we looked at the impacts of regional salinity and temperature, local wave exposure, and site-dependentfacilitation effects on the associated community composition. Wave exposure was the best predictor forspecies assembly structure, which was also affected byMytilus biomass and by salinity and water tempera-ture. While the tested variables provided robust explanations for community structure and density, theydid not provide conclusive explanations for variation in species richness or evenness. Mytilus biomass hada stronger effect on the associated community with increasing wave exposure at the deeper depth, but thepatterns were less obvious at the shallower depth. The latter was also the case for Fucus. These findingscomply partly with theoretical predictions suggesting stronger facilitation effects in physically harsh envi-ronments. Our results indicate that environmental drivers are the main structuring forces that affect speciesassembly structure, but also foundation species are important. Thus, predicting changes in species distribu-tions and biodiversity requires the simultaneous consideration of environmental gradients, as well as thestructure and composition of foundation species and the interplay between these factors. This workadvances our understanding of the processes that modulate species distributions in a marginal marine areaand broadens the knowledge of how biological and environmental factors interplay and have an influenceon hard-bottom community structure in brackish water seas.

Key words: Baltic Sea; biodiversity patterns; community structure; environmental gradients; facilitation; foundationspecies; scale dependency; stress-gradient hypothesis.

Received 30 April 2019; revised 1 August 2019; accepted 5 August 2019. Corresponding Editor: Hunter Lenihan.Copyright: © 2019 The Authors. This is an open access article under the terms of the Creative Commons AttributionLicense, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.� E-mail: [email protected]

❖ www.esajournals.org 1 October 2019 ❖ Volume 10(10) ❖ Article e02883

info:doi/10.1002/ecs2.2883

info:doi/10.1002/ecs2.2883

http://creativecommons.org/licenses/by/3.0/

INTRODUCTION

Foundation species (FS), such as many trees,shrubs, corals, bivalves, macroalgae, and sea-grasses, are species that often have a dispropor-tionately large influence on community structureand ecosystem function (Dayton 1975). They arevital for many systems as they have a propensityto increase diversity, biomass, or ecosystem sta-bility through, for example, enlarging nichespace (Bulleri et al. 2016), enhancing habitatcomplexity (Kostylev et al. 2005), reducing phys-ical stress (Bertness and Callaway 1994), support-ing gas exchange (Attard et al. 2019), andincreasing resource availability (Norkko et al.2006, Norling and Kautsky 2007). High diversityis important, as rich communities are betterequipped with functions that buffer againstfuture changes in the environment, caused bystressors such as climate change, species intro-ductions, or eutrophication. The communityeffects of FS depend on the environmental set-tings and are often conditioned by their individ-ual- or population-level properties (density, age,size). Net effects of FS may therefore vary frompositive, neutral to negative (e.g., reviewed inBateman and Bishop 2017). The stress-gradienthypothesis (SGH; Bertness and Callaway 1994)suggests that the strength of facilitation and com-petition should be inversely related. At the high-stress end, the effect of facilitation on speciescomposition should be strong as facilitation alle-viates environmental stress allowing a higherdiversity or abundance of the associated commu-nity. Yet, mainly due to competition effects,where FS may monopolize resources, interac-tions may be neutral or even negative in physi-cally low-stress environments (Bruno et al. 2003,Crain and Bertness 2006). The influence of FSshould increase with the degree to which theyalleviate stressful conditions. Diversity maytherefore peak in regions where habitat-formingspecies are most abundant and where physicalstructure is a limiting resource (Kimbro andGrosholz 2006). Thus, many factors affect thefacilitating mechanisms of habitat-formingspecies and the challenge is to identify circum-stances under which these habitat-forming spe-cies alleviate stress and provide facilitatingservices to the community (Crain and Bertness

2006). Two decades ago, Bertness et al. (1999)called for a more diversified research related tobiogenic habitat provision. Even if there has beena renaissance in studies related to positive inter-actions since then, facilitation as a mechanismremains understudied in many systems (He andBertness 2014, Wright and Gribben 2017). More-over, while most studies have focused on pair-wise species interactions (Brooker et al. 2008),there has been a call for studies on how facilita-tion affects whole communities (Soliveres et al.2015, Wright and Gribben 2017) and how co-occurring or adjacent habitat-forming species inconjunction determine the species assemblage(Yakovis et al. 2008, Bishop et al. 2013).Apart from facilitation, many environmental

drivers affect species distributions that ultimatelyset the local and regional species pool. These dri-vers occur in spatial hierarchies, they are con-nected, and they integrate the interplay ofecological processes that form the patterns weobserve (Hewitt et al. 2017). In the Baltic Sea, forexample, sea salinity forms an upper-level driverthat ultimately determines how deep into theBaltic Sea marine species penetrate (Remane1934). Another upper-level regional driver istemperature that may determine how southcold-water species are established. A local driveris formed by, for example, wave exposure thatmay affect focal response variables, such as spe-cies interactions. Ecological interactions, how-ever, often cross levels of organization or scales,and the interplay of processes at different scalesmay be much more complicated than first antici-pated (Soranno et al. 2014, Hewitt et al. 2017).Consequently, the inclusion of multiple scales inecological studies is important in order to testpredictions about the influence of processes onecological patterns, as the effects of these pro-cesses may be highly context-dependent (Hewittet al. 2017). Limiting the scope and focusing ononly on a narrow set of scales may also lead tofalse conclusions (Hewitt et al. 2017).Here, we examine the community structure

along wave exposure gradients in three nearbyregions that are characterized by slightly differ-ent salinity conditions. In these regions, twohabitat-forming species live at the limit of theirlower salinity tolerance (Vuorinen et al. 2015).Specific interest lies in the analysis of changes in


WESTERBOM ET AL.

benthic species composition in rocky shore habi-tats dominated by the brown macroalga Fucusvesiculosus and the blue mussel Mytilus trossulus(henceforth Fucus and Mytilus). Both Fucus andMytilus are dominant space occupiers in the areaand provide several important ecosystem ser-vices including supporting and regulating ser-vices. Mytilus has a central role in the Baltic Seafood web. Blue mussels circulate nutrients andenhance biodiversity both directly, via substrateprovision, and indirectly, by affecting near-bedhydrodynamics, affecting sedimentation rates,and clearing the water (Norling and Kautsky2008 with references therein). Fucus spp. is theonly perennial canopy-forming macroalga in thenorthern Baltic Sea and provides living space fora diverse assemblage of species. It also functionsas a spawning, breeding, and foraging groundfor many fish (Saarinen et al. 2018 with refer-ences therein).

While studies generally have looked at singlesystems along gradients of stress or at one-dimensional gradients of disturbance in the samedirection with a non-nested nature, we focussimultaneously on two environmental gradientsand two FS and ask if their facilitating functionsshow responses to changes in the environment.In this rocky shore system, the importance ofmussels and seaweeds as FS has experimentallybeen shown at patch scales (Koivisto and Wester-bom 2010). However, it has largely remainedunclear how different environmental gradientsaffect their facilitating functions. Nor have stud-ies of scale-dependent biodiversity been carriedout in a hierarchically balanced way, coveringentire wave exposure gradients.

The purpose of this study is to examine anddefine the sublittoral hard-bottom community interms of its structure and composition along twonested stress gradients—a regional and physio-logical salinity gradient and a local and physicalwave exposure gradient. Our specific hypotheseswere as follows: (1) Community structure andcomposition differ along environmental gradi-ents and show responses that relate to the quan-tity of the two habitat-forming FS. Salinity is ahighly important driver in this ecosystem withlarge effects on the resident Mytilus population,which shows evident structural changes whenmoving from west to east toward decreasingsalinity (Westerbom et al. 2002, 2019). Based on

prior experimental work (Koivisto and Wester-bom 2010), and knowledge of the regional differ-ences in Mytilus population structure andbiomass (Westerbom et al. 2002, 2019), weexpected to see the most abundant community inthe west, where the biomass of Mytilus is highestand the bed structure is the most complex. Simi-larly, we expected to see a less dense communityin the east, where the biomass of Mytilus is thelowest within the region. Since differences inFucus biomass were not expected among regions,as Fucus is less responsive to salinity changes, wedid not expect to see any region-wide effects ofFucus. Similarly, as the structure and biomass ofthe habitat-forming species vary among waveexposures (Westerbom and Jattu 2006), thesechanges in structure and composition wereexpected to be seen also in the local residentcommunities. According to our second hypothe-sis (2) the facilitating effects of the FS (measuredas high diversity and/or abundance of associatedfauna) are most obvious at sites with high waveexposure, thus supporting the stress-gradienthypothesis (SGH; Bertness and Callaway 1994).

METHODS

Study area and speciesThe Baltic Sea is among the largest brackish

water areas in the world offering a physicallydemanding, low-saline, temporally ice-covered,non-tidal, and evolutionarily young environ-ment. The species pool is small, but species oftenoccur in great density. The coastline of the studyarea, the western Gulf of Finland, is dominatedby sublittoral rocks and reefs that typically arecovered by dense blue mussel beds extend-ing from the water surface down to more than30 m depth with densities sometimes exceeding105 individuals/m2 (Westerbom et al. 2008).Fucus is the largest perennial alga in the studyarea, and it is typically belt-forming at shallowdepths or organized in patches maintained by acombination of availability of light, ice abrasion,herbivory, and competition (Korpinen et al. 2007,Kraufvelin et al. 2007). Gradients in salinity, andto a lesser extent temperature, characterize theGulf of Finland (e.g., Engstr€om-€Ost et al.2015, Westerbom et al. 2019). Sea temperatureincreases slightly toward the east, whereas salin-ity declines from west to east. Even if the decline


WESTERBOM ET AL.

in salinity is subtle within the focal area (from~6.0 to 5.6), it is reflected in the Mytilus popula-tion as higher stability in the west where alsobigger mussels are found compared to the east,where populations are much more dynamic(Westerbom et al. 2019). In addition to salinity,wave exposure is a major structuring force affect-ing local distribution patterns (Westerbom andJattu 2006). Fucus populations have not been sys-tematically surveyed in a comparable way in thisregion, but the coverage of Fucus is approxi-mately similar among regions. Within regions,the coverage of Fucus is typically higher at inter-mediately exposed sites compared to very shel-tered or to very wave-exposed sites.

SamplingUsing identical sampling methods, we sam-

pled mussel bed communities and macroalgalcommunities in three regions along a 60-kmstretch of coastline in SW Finland (approximately59°450N, 22°400–59°470N, 23°420). These regionswere Hanko in the west, Tv€arminne in the center,and Jussar€o in the east (Fig. 1). We collected sam-ples from 36 stations (12 per region) including arange of wave exposures where stands of Fucusand beds of Mytilus occur. To estimate waveexposure, we used the Baardseth wave exposureindex (BI; see Westerbom and Jattu 2006), andconsistent with recommendations of He and

Bertness (2014), we grouped the stations intofour exposure classes (sheltered [s], BI = 0–2;moderately sheltered [ms], BI = 7–9; moderatelyexposed [me], BI = 13–15; and exposed [e],BI = 18–22). Each exposure level included threestations per region (Fig. 1). Because shore slopeaffects the distribution of habitat-forming mus-sels (Westerbom et al. 2008), we chose onlyislands with a smooth rocky slope. We executedthe sampling hierarchically, with samples nestedwithin stations and stations nested within expo-sure levels and regions.Within each station, we took ten samples, five

from each of the two depths, 3 and 8 m, at dis-tances of 10–20 m within depths (total number ofsamples, N = 360). At each station in August–September, we collected benthic samples fromslopes opening toward the highest wave expo-sure, using a 20 9 20 cm frame with a 0.1-mmnet bag. Samples were systematically dispersedin time within regions and exposures to avoidtemporal biases in sampling. The method is stan-dard in the Baltic Sea and used in, for example,national monitoring programs of the zoo- andphytobenthos (Kautsky and Van der Maarel1990, Westerbom et al. 2002, R�aberg and Kaut-sky 2007, Wikstr€om and Kautsky 2007). Then, weprocessed the samples in the laboratory andcounted the species according to common prac-tices (Koivisto and Westerbom 2010, D�ıaz et al.

me

meme

msms

ms

s

e

ee

s s

s

msmememe

e e e

sss

ms msmeme me

e e e

HANKO TVÄRMINNE JUSSARÖ

10 km

ms ms ms

ss

59⁰ 50

22⁰ 50

Fig. 1. Map of the study area. Stations shown with letters, where s denotes sheltered, ms moderately sheltered,me moderately exposed, and e exposed.


WESTERBOM ET AL.

2015). Biomasses for Mytilus were estimatedaccording to Westerbom et al. (2002). Fucus andother algae in samples were dried at 60°C for 5 dand then weighed.

In addition to sampling, we lowered a 100-mtransect line perpendicularly from the shorelineand stretched it out to a minimum depth of15 m. Once in place, a scuba diver estimated thecoverage of flora and fauna over an area of 4 m2

on both sides of the transect for every depthmeter. Then, we extracted transect data from 3and 8 m depths for use in multivariate models asmeasures of seascape structure. The coverage ofFucus, perennial algae, and red algae was used asa variable for seascape structure. To account forspatial variability and the lack of environmentalmonitoring data from each of the 36 stations, weextracted environmental data from coastal datamodels that are based on Finnish governmentalin situ monitoring or satellite data and hosted bythe Finnish Environment Institute. These dataincluded sea surface salinity, water transparency(i.e., Secchi depth), phytoplankton chlorophyll a(Chl-a), and sea surface temperature.

Data analysisWe applied parametric univariate techniques

to the facilitator variables (Mytilus biomass andabundance, Fucus biomass) and to summaryvariables of the associated macrofauna commu-nity (total abundance, Margalef species richness,Pielou’s evenness) to test for differences amongregions, wave exposure levels, and sampling sta-tions. Due to the broad difference between mus-sel and seaweed habitat structure and the lack ofFucus at 8 m depth, we ran the analyses sepa-rately for 8 m depth (only facilitation by Mytilus)and for 3 m depth (Fucus and Mytilus facilita-tion).

For univariate analyses of the facilitator vari-ables and summary variables for the associatedmacrofauna, we used three-way nested mixed-model ANOVAs (Underwood 1997) followingthe model:

Xijkl ¼ lþ Ri þ Ej þ SðREÞkðijÞ þ REij þ elðijkÞ

where Ri denotes level i of the region (fixed), Ej

denotes level j of exposure level (random), and S(RE)k(ij) denotes level k of station (random) nestedin exposure level j and region i. REij representsan interaction component of the ith level of factor

region and the jth level of factor exposure level,and el(ijk) indicates the difference between anindividual sample at station k of combination iand j of region and exposure level, respectively(Underwood 1997). Wave exposure was treatedas random factor because the used wave expo-sure values represent a continuum of exposurelevels rather than clearly separated classes (seeUnderwood 1997). We used a posteriori Student-Neuman-Keuls (SNK) tests to reveal the occur-rence and direction of pairwise differences.We studied facilitation effects with generalized

additive mixed models (GAMMs) since theyopen the possibility to examine non-linear rela-tionships (Wood 2006). We selected one modelfor each of the two depths to describe the varia-tion in abundance of associated individualsbetween the two setups of habitat-forming FS.Initially, the factors used in each model wereregion (three levels) and wave exposure (fourlevels), both representing gradients of stress anddisturbance, respectively. Since region had noeffect, the final model excluded region as a factor.We considered the station factor as a randomeffect following the guidelines listed by Zuuret al. (2010). In the first model (8 m), Mytilus wasthe only foundation species, while in the secondmodel (3 m), the factor for FS (BiomassFS) con-tained two levels, the biomasses of Fucus andMytilus. On the continuous variable (abundance),we applied a smoother to counteract for residualheteroscedasticity due to non-linearity in thefacilitation relationships. In the 3 m model, weused one smoother for each of the two FS. Thedependent variable for associated fauna, that is,abundance of associated individuals, was dis-crete and over-dispersed showing a negativebinomial distribution, and no transformationwas applied to this variable. We applied themodel selection following the criterion estab-lished by Zuur et al. (2010) on which outlierswere identified and eliminated. We examinedcollinearity by plotting covariates together, whilewe detected variance inflation by the VIF index(Zuur et al. 2010). Collinearity was not an issuein these models. In both models, we used a cubicspline smoother (Zuur 2012) and we controlledoverfitting using the integrated square secondderivative cubic spline penalty (Wood 2006, Zuur2012). Then, we used backward AIC selection(AIC, Akaike information criterion) to find the


WESTERBOM ET AL.

optimal models (Zuur et al. 2010). Finally, weapplied model validation to verify that therewere no violations of heteroscedasticity. Specifi-cally, we plotted the relationship between residu-als and fitted values. Afterward, we plottedresiduals against each co-variable to investigatepossible model misfit. The models we used wereas follows:

Model 8m: Abundance indijk � Interceptþ smootherðMytilus BiomassixExposurejÞþ Random effectsðstationskÞ

Model 3m: Abundance indijkl � Interceptþ Exposurej þ smoother

ðBiomassFSixFoundation specieskÞþ Random effectsðstationslÞ

To visualize and to test for differences in com-munity structure, we ran nonparametric multi-variate techniques (Anderson et al. 2008) on theentire macrofauna data set applying the samemodel structure as for the three-way nestedmixed-model ANOVAs above. Using three-waynested PERMANOVA, separately for the twostudy depths, we thus tested the data for differ-ences in community structure among the threeregions, four exposure levels, and 36 samplingstations. To even out the relative influence fromvery dominant and more rare species in the mul-tivariate analyses, we square-root-transformedthe data before running Bray-Curtis similarityanalyses. Furthermore, we analyzed the associa-tion between the observed species compositionpattern and the registered environmental vari-ables using distance-based linear modeling,DISTLM, separately for each depth and usingnormalized environmental variables. Pearson’scorrelation coefficient indicated low correlationbetween environmental variables. We ran theDISTLM analyses using the “Best” option withinthe PRIMER software, which examines all possi-ble combinations of predictor variables withinthe PRIMER software (Clarke and Gorley 2006,Anderson et al. 2008). Within this process, weinitially obtained the best models for a series ofalternative models including an increasing num-ber of explanatory variables, going from a modelincluding only one variable and ending with a

model including all potential predictor variables.Then, we identified the most parsimoniousmodel among these by AIC and we visualizedthe final model by distance-based redundancyanalysis, dbRDA, which is an ordination tech-nique constrained to find linear combinations ofpredictor variables that explain the greatest vari-ation in the data cloud (Anderson et al. 2008).We assessed the relative influences of eachresponse variable and environmental variablebased on the length of their overlaid vectors inthe resulting ordination plot. In addition, weidentified the individual relationship of eachenvironmental variable to the observed patternin the species data set based on marginal F tests.We ran the initial univariate analyses in

GMAV5, the GAMM analyses in R version 3.4package mgcv, and the multivariate statisticalanalyses in PRIMER 7.0 and PERMANOVA+ forPRIMER. Before running the parametric univariatetests, we checked the normality by Kolmogorov-Smirnov’s test and homogeneity of variances byCochran’s C test. To homogenize variances, weused a square-root or logarithmic transformation.

RESULTS

Distribution of Mytilus and Fucus along gradientsand their effects on faunaAt 8 m depth, there was a significant interaction

effect (P = 0.01) between region and exposure forMytilus biomass (Fig. 2A; Appendix S1: Table S1a).Hanko showed higher Mytilus biomass thanTv€arminne and Jussar€o in the three lowest expo-sure levels, while for the highest exposure level,Tv€arminne showed the highest biomass. At Hankoand Tv€arminne, there were also significant within-region differences (Fig. 2A), whereas at Jussar€o, noexposure differences were seen.Mytilus abundanceshowed no region-wide effects, although therewere highly significant exposure level effects(P < 0.001) with abundances generally increasingwith increased level of wave exposure (Fig. 2A;Appendix S1: Table S1A).At 3 m depth, a significant interaction

(P < 0.01) between region and exposure was seenfor Mytilus biomass (Fig. 2B; Appendix S1:Table S1b) arising from the very high biomassesat sheltered shores in the Hanko region and noexposure differences within the other tworegions (Fig. 2B). Mytilus abundance showed


WESTERBOM ET AL.

Fig. 2. Distribution of the predictor variables Mytilus biomass, Mytilus abundance, and Fucus biomass and theresponse variables richness, evenness, and total abundance of associated macrofauna for (A) 8 m depth and (B)3 m depth. Note differences in scale among y-axes. Bars are means + SE. S stands for sheltered, MS for moder-ately sheltered, ME for moderately exposed, and E for exposed.


WESTERBOM ET AL.

neither region nor exposure effects. Fucusbiomass at 3 m depth showed no regional differ-ences, but showed wave exposure effects(P = 0.013), with the medium-sheltered region(ms) presenting higher biomasses than the otherthree exposure levels averaged over the regions(Fig. 2B; Appendix S1: Table S1b).

In total, we identified 52 invertebrate taxa ortaxon groups associated with Mytilus or Fucus.At 8 m depth, the total number of individuals ofassociated fauna showed a strong response tothe interaction between region and exposure(P < 0.001) that largely followed the results ofMytilus biomass (Fig 2A; Appendix S1:Table S1c) and varied with exposure (P = 0.005).Species richness showed a small effect of regionat 8 m (P = 0.02), due to higher richness at Jus-sar€o compared to Tv€arminne and Hanko. No dif-ferences in evenness were seen among regions,nor among exposure levels (Appendix S1:Table S1c). At 3 m depth, the total abundance ofassociated macrofauna showed a weak exposurelevel effect that was due to constantly higherabundances at the most sheltered stations com-pared to the other wave exposure levels(P = 0.02; Fig. 2B; Appendix S1: Table S1d). Noregional effect on richness was seen, but richnesswas higher at the two lower exposure levels thanat the other two levels (P = 0.002). Evenness wasslightly lower at Hanko compared to the otherregions (P = 0.022; Appendix S1: Table S1d). Dif-ferences in evenness were, however, very smallwithin and among regions (Fig. 2B).

Facilitation effects along stress gradientsWave exposure had a clear effect on the rela-

tionship between the biomass of Mytilus and theabundance of associated species. This relation-ship was non-linear at both depths, showing thatthe facilitating function of Mytilus increasedat higher Mytilus biomasses (P ≤ 0.001 in bothcases). At 8 m, the facilitating effect increasedsigmoidally with increasing biomass of Mytilus(Appendix S2: Table S1a). The facilitating effectwas strongest at the most exposed stations andapproached an asymptote at all exposures, indi-cating that Mytilus had reached a biomass thatmaximized its facilitating potential (Fig. 3A;Appendix S2: Table S1a). At 3 m depth, the effectof the facilitating function of Mytilus resembledthe exponential function with stronger positive

effects at sheltered and exposed stations than atintermediate stations (Fig. 3B; Appendix S2:Table S1b). The smoother on Fucus showed thatwhile there was a positive linear relationshipbetween the biomass of Fucus and the density ofthe associated community, the importance ofFucus appeared not to change along thewave exposure gradient (Fig. 3B; Appendix S2:Table S1b).

Associated fauna: multivariate responsesThe community structure of associated

macrofauna at 8 m depth showed a significantinteraction effect between region and exposure(P < 0.01). There were also highly significant dif-ferences among all stations nested in region andexposure (P < 0.001; Appendix S3: Table S1). At3 m depth, the structure of the associated macro-fauna community differed significantly amongexposure levels as well as among stations(P < 0.001), but no differences occurred amongregions. Pairwise tests revealed that the differ-ences between exposure levels occurred betweenall pairs, except for the two intermediate (ms andme) and the two outermost levels (me and e). At8 m depth (with only Mytilus as facilitator), weidentified the most parsimonious DISTLM modelfor a combination of nine environmental vari-ables with wave exposure (21%), Mytilus abun-dance (20%), temperature (14%), salinity (13%),and Mytilus biomass (13%) as the strongest pre-dictors (Fig. 4; Appendix S4: Table S1). Samplesfrom different exposure levels grouped distinctlyand systematically along the dbRDA1 axis in theordination (Fig. 4). Mytilus abundance, waveexposure, and salinity were strong predictors forpositive values along the dbRDA1, while temper-ature predicted negative values. The samplesfrom different regions were largely groupedalong the dbRDA2 from top (mainly Hanko) tobottom (mainly Jussar€o). Mytilus biomass was astrong predictor of positive values along thedbRDA2, while the transect coverage of redalgae was a strong predictor of negative values.The taxa showing the strongest influence on theobserved patterns were amphipods of the genusGammarus that demonstrated a positive relation-ship with wave exposure and the bivalve Myaarenaria, which correlated negatively with waveexposure (Fig. 4). The second axis was mainlyinfluenced by the gastropod Hydrobia, which


WESTERBOM ET AL.

Fig. 3. Results from the GAMM showing differences between the FS Mytilus and Fucus and their effects on thetotal community associated across a gradient of wave exposure. (A) Results from 8 m depth for Mytilus. (B)Results from 3 m depth for both Mytilus and Fucus. S stands for sheltered, MS for moderately sheltered, ME formoderately exposed, and E for exposed.


WESTERBOM ET AL.

Fig. 4. Distance-based redundancy analysis (dbRDA) for square-root-transformed species abundance datafrom 8 m depth with the most important discriminating environmental/background variables and species as vec-tor overlays. As shown, different regions (along the y-axis) and different exposures (along the x-axis) appear indifferent ends of the cluster, largely paralleling differences in Mytilus bed structure among regions and expo-sures. In the legend, H stands for Hanko, T for Tv€arminne, and J for Jussar€o, while s stands for sheltered, ms formoderately sheltered, me for moderately exposed, and e for exposed.


WESTERBOM ET AL.

showed a positive relationship with Mytilus bio-mass.

We identified the most parsimonious model at3 m depth (with both Mytilus and Fucus as facili-tators) for a combination of 12 environmentalvariables. Marginal tests identified wave expo-sure (22%) as the single variable that was moststrongly attributed to changes in the associatedmacrofauna community, followed by tempera-ture (16%) and salinity (7%; Appendix S5:Table S1). Fucus and Mytilus, however, explained

only a fraction of the variation (2% and 4%,respectively). As Fig. 5 shows, samples from dif-ferent exposure levels were grouped along thedbRDA1 with samples from sheltered stations tothe right and samples from increasingly moreexposed stations further to the left, the latter alsocorrelating positively with salinity andnegatively with temperature. With regard todbRDA2, this axis encompassed a considerablysmaller share of the explained variation with bio-mass and abundance of Mytilus as predictors of

Fig. 5. Distance-based redundancy analysis (dbRDA) for square-root-transformed species abundance datafrom 3 m depth with the most important discriminating environmental/background variables and species as vec-tor overlays. As shown, there is a high similarity among regions (y-axis), but clear differences among wave expo-sure levels (x-axis). In the legend, H stands for Hanko, T for Tv€arminne, and J for Jussar€o, while s stands forsheltered, ms for moderately sheltered, me for moderately exposed, and e for exposed.


WESTERBOM ET AL.

positive values and transect coverage of redalgae as a predictor of negative values. The taxashowing the strongest influence on the observedpatterns were Gammarus spp. that showed apositive relationship with wave exposure, andthree species of gastropods together withMacoma balthica showing negative relationshipswith wave exposure (Fig. 5).

DISCUSSION

A peculiar feature of the rocky sublittoral inthe Baltic Sea is that species diversity on localand landscape scales shows relatively small vari-ations. Quality differences within and betweenrocky habitats in the Baltic Sea are typicallyexpressed in dominance or abundance shiftsamong species, rather than in richness. The smalldifference in terms of biodiversity in this study,within and among regions and exposure levels,is likely a result of this omnipresent nature ofspecies and an overall small species pool. Thelarge-scale environmental variables salinity andtemperature appear to cause region-wide effects,displayed as distinct groupings of samples fromthe different regions in the sample cloud of thedbRDA, with regions falling in a logical longitu-dinal order (Fig. 4). However, this distinctivepattern only characterizes samples from 8 mdepth, whereas samples from 3 m depth do notreflect any distinct patterns driven by gradientsin salinity or temperature. The effects of Mytilusare also clearly displayed in the dbRDA cloud(Fig. 4), where samples with lowest Mytilus bio-mass at Hanko and especially samples with high-est biomass at Tv€arminne (Fig. 2) are groupedapart from the main regional cloud, character-ized by significantly more (Hanko) or signifi-cantly less (Tv€arminne) Mytilus biomass. Thesefindings from 8 m depth support Hypothesis 1—predicting community differences driven by theinterplay of regional-scale and local-scale gradi-ents. Samples from 3 m depth show patterns thatare driven by predominantly local drivers, sug-gesting that the local wave exposure gradientcompletely overrides the imprint of regionalfactors.

This paper demonstrates that faunal composi-tion and faunal abundance in our study areashow distinct differences that are (1) influencedby large-scale abiotic factors including salinity

and temperature (supporting Hypothesis 1), (2)locally modified by wave exposure (supportingHypothesis 2), and (3) driven by the facilitatingfunction of Mytilus (supporting Hypothesis 2)while being less affected by Fucus (supportingweakly Hypothesis 1). With regard to the stress-gradient hypothesis (Bertness and Callaway1994), Mytilus at 8 m depth clearly affects theassociated community and increases the totalabundance of associated fauna with strongereffects at exposed stations compared to shelteredones (supporting Hypothesis 2; Fig. 3A). How-ever, at the higher wave exposure stress at 3 mcompared to 8 m depth, the pattern is less obvi-ous and indicates that the facilitation function, infact, may be strongest at the most sheltered sta-tions (see also D�ıaz et al. 2015). The challenge inthis context is therefore to develop an under-standing of how the environment modulates thecommunity structure differently at the twodepths and how wave exposure changes theinteraction between the invertebrate communityand the two habitat-forming species. The differ-ence in community response between depthsand among exposures may arise from two inter-related factors: (1) those of how increasing physi-cal stress changes the intensity or function of thefacilitators affecting their facilitating value,and (2) those arising from depth-dependent dif-ferences in the dominance and heterogeneity ofdifferent habitats. We address these two possibil-ities below.Many studies suggest a decline in facilitation

toward the most severe conditions, as facilitatorsare less successful in ameliorating stress in veryharsh environments (Kawai and Tokeshi 2007;but see He and Bertness 2014, Qi et al. 2018).Therefore, while the facilitation of mussels at8 m depth increases with exposure throughoutthe wave exposure gradient (Fig. 3A), musselsare unable to ameliorate stress at the more stress-ful physical conditions that are prevailing at 3 mdepth at the exposed stations (Fig. 3B). Thiseffect may arise from the interacting effects ofmussel bed structure and wave force strengththat differs among stations and between depthswithin stations. Such effects may result fromenvironmental effects on the facilitating traits ofthe FS (Irving and Bertness 2009, Bishop et al.2013, Bateman and Bishop 2017). At 3 m depth,mussel beds are mainly composed of many small


WESTERBOM ET AL.

individuals that are tightly packed and have aless complex bed structure but more patchy dis-tributions (D�ıaz et al. 2015) than at 8 m (see theratio between abundance and biomass at differ-ent exposures in Fig. 2A, B). This difference inmussel bed structure is driven by wave-induceddislodgement that increases isometrically withincreasing size of mussels and affects beds at 3 mdepth significantly more than at 8 m depth(Westerbom and Jattu 2006). Such loss of archi-tectural complexity provided by habitat-formingmussels and oysters has been shown to reducebiodiversity, species richness, or evenness (Kosty-lev 1996, Gutierrez et al. 2003, Kimbro andGrosholz 2006, Koivisto and Westerbom 2010).Results here suggest that the bed structure ofMytilus at 8 m depth offers a more complexthree-dimensional architecture throughout thegradient, whereas a similar structure at 3 m isonly found at sheltered stations where the waveforce is not sufficiently strong to remove the lar-gest individuals (Westerbom and Jattu 2006).

An alternative explanation may arise from thedifference in physical structure between depths,as bed surface area may be a stronger predictorfor individual abundance and species richnessthan environmental stress (Kostylev et al. 2005,Kimbro and Grosholz 2006). Large musselsincrease the available surface area for fouling andhiding organisms, and this surface area may belargest at low-disturbance stations where musselsare larger. Additionally, the degree of shell pack-ing and the size of mussels affect refuge provision(Kostylev et al. 1997, Commito and Rusignuolo2000) and affect further the enrichment of organicmaterial and nutrients, which typically are limit-ing resources on wave-beaten shores (Tokeshi andRomero 1995, Norling and Kautsky 2008, Kottaet al. 2009). FS may have a particularly largeinfluence on associated communities where theyincrease the availability of limiting resources(Bateman and Bishop 2017). At highly wave-bea-ten shores at 3 m, the availability of these musselmatrix resources wanes directly throughincreased flushing and indirectly through packingthe mussel beds and thereby reducing their sedi-ment storage potential.

A third explanation may arise from depth-dependent differences in the dominance struc-ture of habitats. Mussels below the algal belt areprimary space holders that almost solely provide

habitats and resources for secondary species. Inthe windswept Patagonian intertidal, Sillimanet al. (2011) showed that competition effects bymussels become unimportant as mussels are thesole dominant space holders capable of bufferingagainst lethal conditions, and therefore, thewhole community is completely dependent onmussels for its persistence. Similarly, Norling andKautsky (2008) and Koivisto and Westerbom(2010) showed experimentally that mussel bedshost substantially more species and individualsthan non-mussel habitats in the Baltic Sea. McA-fee et al. (2016) showed among Australian oys-ters that the effects of facilitators may bedependent on the extent to which facilitatorsreplicate the function of other co-occurring FS.Hence, at 8 m depth where there are no alterna-tive habitat-forming biogenic structures, speciesrichness and individual density may be particu-larly dependent on the availability of morehomogeneous mussel beds (D�ıaz et al. 2015).However, at shallow depths, mussel beds are notthe sole biogenic habitat providers (Kraufvelinand Salovius 2004), likely deflating the impor-tance of one habitat provider (Saarinen et al.2018). In a similar way, two habitat-forming mus-sel species coexist and facilitate each other in themid-zone of rocky shores on the South Africansouth coast, while on the lower and upper zones,either species dominates due to contrasting spe-cies-specific tolerances to wave and heat stress(Erlandsson et al. 2011).A corresponding mechanism driven by the

amelioration of wave exposure was not found forFucus. Fucus explained only a fraction of the com-munity structure (Appendix S1: Table 4). Thecommunity effects of canopy-forming specieshave been widely studied in the intertidal wherecanopies provide shelter from temperature stress(Watt and Scrosati 2013). Under those conditions,fucoids have shown positive effects on understoryorganisms in physiologically harsh and physicallybenign environments by providing shelter fromheat, whereas negative effects have been shownin physiologically benign and physically harshenvironments as algal fronds smash and scourtheir immediate surrounding due to waves (Beer-mann et al. 2013 with references therein). In theshallow subtidal, disturbance caused by the thal-lus movement can be very pronounced and accen-tuate with higher wave exposure. Kiirikki (1996),


WESTERBOM ET AL.

for example, showed that Fucus controls the vol-ume of filamentous algae on exposed shores, buton sheltered shores, its effect was negligible. Algalsmashing and scouring may therefore reduce totalindividual abundance directly, but also indirectlyby reducing Mytilus coverage or the size of mus-sel individuals in the patches below the Fucus(Appendix S6: Fig. S1). One foundation speciesmay therefore affect the traits and the facilitatingfunction of the other (Bishop et al. 2013). In addi-tion, many other plausible explanations may clar-ify the observed patterns. For instance, whenmultiple stressors work in the same direction, thepredictability may suffer, as species may be toler-ant to one stressor and intolerant to another (Qiet al. 2018). Moreover, facilitation as a mecha-nism, as other biotic interactions, is primarily spe-cies-specific and responses may vary fromnegative to positive among species due to theirdifferent ecological requirements (Soliveres et al.2015).

The species community and diversity patternsobserved in this study are much in line with pre-vious research efforts in the area. These showweak effects of Mytilus on species richness, butlarge effects on invertebrate abundance and com-munity structure (Norling and Kautsky 2008,Koivisto and Westerbom 2010, Koivisto et al.2011), as well as greater facilitation effects at 8 mthan at shallower depths (D�ıaz et al. 2015). Theoverall weak facilitator effect of Fucus across dif-ferent wave exposures is surprising, as Fucus tra-ditionally has been considered as an importantFS in the Baltic Sea (R�aberg and Kautsky 2007,Wikstr€om and Kautsky 2007) and has experi-mentally been shown to increase diversity andindividual density (Koivisto and Westerbom2010). Nevertheless, previous studies have beenrather inconclusive when comparing Fucus withephemeral algal habitats. Some have shownhigher density of the associated community inFucus habitats, but similar diversity (Kraufvelinand Salovius 2004, Wikstr€om and Kautsky 2007),while others have shown lower density buthigher diversity (Saarinen et al. 2018). Theseinconclusive results suggest that the positivecommunity effects of Fucus are highly context-dependent.

Even if the influence of wave exposure on Bal-tic Sea communities has been recognized for long(Kautsky and Van der Maarel 1990) and is also

supported by an ample global literature(McQuaid and Branch 1984, Menge and Suther-land 1987, Harley and Helmuth 2003), manystudies in the Baltic Sea still overlook the impactof wave exposure as a determinant of commu-nity composition. We show that apart frominfluencing on community composition, waveexposure seems to affect the interactions betweenspecies. Therefore, studies focusing on the habi-tat-forming function of aquatic organisms needto consider the influence of wave exposure thataffects the community not only directly, but alsoindirectly through changing the facilitating traitsof the habitat-forming species. Failure to inte-grate site-specific interaction reduces our abilityto predict community responses to, for example,a changing climate and falls short in identifyingthe generality of the mechanisms behind the fac-tors that affect the structure of communities (Gil-man et al. 2010). Overall, we need a betterunderstanding of the mechanisms that affect therelationships between foundation species, theenvironment within which they interact, and theshared effects these variables impose on the asso-ciated community. The challenge for scientists isto develop accurate predictive models abouthow these factors interplay, especially as globalseas are rapidly changing due to a shifting cli-mate. As other seas, the Baltic Sea shows risingtemperatures, declining salinity conditions, andpossibly an altered wave climate. Changes inthese factors are predicted to shift the distribu-tion and structure of the focal foundation speciesof this study (Vuorinen et al. 2015, Westerbomet al. 2019). Such structural changes among FStogether with an altered sea climate (Westerbomet al. 2019) will likely affect the resident commu-nity (see also Koivisto and Westerbom 2010, D�ıazet al. 2015). Consideration of the structure andimpact of FS along spatial and temporal scales istherefore critical for understanding the underly-ing mechanisms affecting species distributionand for finding management solutions to miti-gate biodiversity loss (Hewitt et al. 2017). Usinga hierarchical sampling design, we have gainedinsights to the ways positive species interactionsvary with environmental conditions and haveempirically shown that facilitation contributes tocommunity structure in a sparsely studied com-munity at the range border. However, using amensurative approach, this study was not able to


WESTERBOM ET AL.

provide any mechanistic explanations for theprocesses underlying the observed patterns. Weunderline the need for manipulative experimentsthat can test mechanistic relationships to furtherour understanding of species distribution pat-terns that are vital for finding efficient manage-ment solutions and conservation strategies.

ACKNOWLEDGMENTS

The work was financed by Walter and Andr�ee deNottbeck Foundation (to Mats Westerbom, Olli Musto-nen) and Academy of Finland, project number 126831(to Mats Westerbom). The authors declare no conflict ofinterest.

LITERATURE CITED

Anderson, M. J., R. N. Gorley, and K. R. Clarke. 2008.PERMANOVA+ for PRIMER: guide to softwareand statistical methods. PRIMER-E, Plymouth, UK.

Attard, K. M., I. F. Rodil, P. Berg, J. Norkko, A.Norkko, and R. Glud. 2019. Seasonal metabolismand carbon export potential of a key coastal habitat:the perennial canopy-forming macroalga Fucus vesicu-losus. Limnology and Oceanography 64:149–164.

Bateman, D. C., and M. J. Bishop. 2017. The environ-mental context and traits of habitat-formingbivalves influence the magnitude of their ecosys-tem engineering. Marine Ecology Progress Series563:95–110.

Beermann, A. J., J. A. Ellrich, M. Molis, and R. A. Scro-sati. 2013. Effects of seaweed canopies and adultbarnacles on barnacle recruitment: the interplay ofpositive and negative influences. Journal of Experi-mental Marine Biology and Ecology 448:162–170.

Bertness, M. D., and R. M. Callaway. 1994. Positiveinteractions in communities. Trends in Ecology andEvolution 9:191–193.

Bertness, M. D., G. H. Leonard, J. M. Levine, P. R.Schmidt, and A. O. Ingraham. 1999. Testing therelative contribution of positive and negative inter-actions in rocky intertidal communities. Ecology80:2711–2726.

Bishop, M. J., J. Fraser, and P. E. Gribben. 2013. Mor-phological traits and density of foundation speciesmodulate a facilitation cascade in Australian man-groves. Ecology 94:1927–1936.

Brooker, R. W., et al. 2008. Facilitation in plant com-munities: the past, the present, and the future. Jour-nal of Ecology 96:18–34.

Bruno, J. F., J. J. Stachowicz, and M. D. Bertness. 2003.Inclusion of facilitation into ecological theory.Trends in Ecology and Evolution 18:119–125.

Bulleri, F., J. F. Bruno, B. R. Silliman, and J. J. Stachow-icz. 2016. Facilitation and the niche: implicationsfor coexistence, range shifts and ecosystem func-tioning. Functional Ecology 30:70–78.

Clarke, K. R., and R. N. Gorley. 2006. PRIMER v.6: usermanual/tutorial. PRIMER-E Ltd, Plymouth, UK.

Commito, J. A., and B. R. Rusignuolo. 2000. Structuralcomplexity in mussel beds: the fractalgeometry of surface topography. Journal ofExperimental Marine Biology and Ecology 255:133–152.

Crain, C. M., and M. D. Bertness. 2006. Ecosystemengineering across environmental gradients: impli-cations for conservation and management. BioS-cience 56:211–218.

Dayton, P. K. 1975. Experimental evaluation of ecologi-cal dominance in a rocky intertidal algal commu-nity. Ecological Monographs 45:137–159.

D�ıaz, E. R., J. Erlandsson, M. Westerbom, and P. Kraufvelin.2015. Depth-related spatial patterns of sublittoralblue mussel beds and their associated macrofaunadiversity revealed by geostatistical analyses. Mar-ine Ecology Progress Series 540:121–134.

Engstr€om-€Ost, J., I. S. Rasic, A. Brutemark, R. Rancken,G. S. Simi�c, and A. T. Laugen. 2015. Can Cylindros-permopsis raciborskii invade the Baltic Sea? Envi-ronmental Reviews 23:161–169.

Erlandsson, J., C. D. McQuaid, and M. Sk€old. 2011.Patchiness and co-existence of indigenous andinvasive mussels at small spatial scales: the interac-tion of facilitation and competition. PLOS ONE 6:e26958.

Gilman, S. E., M. C. Urban, J. Tewksbury, G. W. Gilchr-ist, and R. D. Holt. 2010. A framework for commu-nity interactions under climate change. Trends inEcology and Evolution 25:325–331.

Gutierrez, J. L., C. G. Jones, D. L. Strayer, and O. O.Iribarne. 2003. Mollusks as ecosystem engineers:the role of shell production in aquatic habitats.Oikos 101:79–90.

Harley, C. D., and B. S. Helmuth. 2003. Local-andregional-scale effects of wave exposure, thermalstress, and absolute versus effective shore level onpatterns of intertidal zonation. Limnology andOceanography 48:1498–1508.

He, Q., and M. D. Bertness. 2014. Extreme stresses,niches, and positive species interactions alongstress gradients. Ecology 95:1437–1443.

Hewitt, J. E., S. F. Thrush, and C. Lundquist. 2017.Scale-dependence in Ecological Systems. in eLS.John Wiley and Sons, Chichester, UK. https://dx.doi.org/10.1002/9780470015902.a0021903.pub2

Irving, A. D., and M. D. Bertness. 2009. Trait-depen-dent modification of facilitation on cobble beaches.Ecology 90:3042–3050.


WESTERBOM ET AL.

https://dx.doi.org/10.1002/9780470015902.a0021903.pub2

https://dx.doi.org/10.1002/9780470015902.a0021903.pub2

Kautsky, H., and E. Van der Maarel. 1990. Multivariateapproaches to the variation in phytobenthiccommunities and environmental vectors inthe Baltic Sea. Marine Ecology Progress Series60:169–184.

Kawai, T., and M. Tokeshi. 2007. Testing the facilita-tion-competition paradigm under the stress-gradi-ent hypothesis: decoupling multiple stress factors.Proceedings of the Royal Society of London B, Bio-logical Sciences 274:2503–2508.

Kiirikki, M. 1996. Experimental evidence that Fucusvesiculosus (Phaeophyta) controls filamentous algaeby means of the whiplash effect. European Journalof Phycology 31:61–66.

Kimbro, D. L., and E. D. Grosholz. 2006. Disturbanceinfluences oyster community richness and even-ness, but not diversity. Ecology 87:2378–2388.

Koivisto, M. E., and M. Westerbom. 2010. Habitatstructure and complexity as determinants of biodi-versity in blue mussel beds on sublittoral rockyshores. Marine Biology 157:1463–1474.

Koivisto, M., M. Westerbom, and A. Riihim€aki. 2011.Succession driven facilitation of macrofaunal com-munities in sublittoral blue mussel habitats. MarineBiology 158:945–954.

Korpinen, S., V. Jormalainen, and T. Honkanen. 2007.Effects of nutrients, herbivory, and depth on themacroalgal community in the rocky sublittoral.Ecology 88:839–852.

Kostylev, V. 1996. Spatial heterogeneity and habitatcomplexity affecting marine littoral fauna. Disser-tation. G€oteborg University, G€oteborg, Sweden.

Kostylev, V., J. Erlandsson, and K. Johannesson. 1997.Microdistribution of the polymorphic snail Lit-torina saxatilis (Olivi) in a patchy rocky shore habi-tat. Ophelia 47:1–12.

Kostylev, V. E., J. Erlandsson, Y. M. Mak, and G. A.Williams. 2005. The relative importance of habitatcomplexity and surface area in assessing biodiver-sity: fractal application on rocky shores. EcologicalComplexity 2:272–286.

Kotta, J., K. Herk€ul, I. Kotta, H. Orav-Kotta, and V.Lauringson. 2009. Effects of the suspension feedingmussel Mytilus trossulus on a brackish watermacroalgal and associated invertebrate commu-nity. Marine Ecology 30:56–64.

Kraufvelin, P., A. T. Ruuskanen, N. Nappu, and M.Kiirikki. 2007. Winter colonisation and successionof filamentous algae on artificial substrates andpossible relationships to Fucus vesiculosus settle-ment in early summer. Estuarine, Coastal and ShelfSciences 72:665–674.

Kraufvelin, P., and S. Salovius. 2004. Animal diver-sity in Baltic rocky shore macroalgae: Can Clado-phora glomerata compensate for lost Fucus

vesiculosus? Estuarine, Coastal and Shelf Sciences61:369–378.

McAfee, D., V. J. Cole, and M. J. Bishop. 2016. Latitudi-nal gradients in ecosystem engineering by oystersvary across habitats. Ecology 97:929–939.

McQuaid, C. D., and G. M. Branch. 1984. Influence ofsea temperature, substratum and wave exposureon rocky intertidal communities: an analysis of fau-nal and floral biomass. Marine Ecology ProgressSeries 19:145–151.

Menge, B. A., and J. P. Sutherland. 1987. Commu-nity regulation: variation in disturbance, competi-tion, and predation in relation to environmentalstress and recruitment. American Naturalist130:730–757.

Norkko, A., J. E. Hewitt, S. F. Thrush, and G. A.Funnell. 2006. Conditional outcomes of facilitationby a habitat-modifying subtidal bivalve. Ecology87:226–234.

Norling, P., and N. Kautsky. 2007. Structural and func-tional effects ofMytilus edulis on diversity of associ-ated species and ecosystem functioning. MarineEcology Progress Series 351:163–175.

Norling, P., and N. Kautsky. 2008. Patches of the mus-sel Mytilus sp. are islands of high biodiversity insubtidal sediment habitats in the Baltic Sea. Aqua-tic Biology 4:75–87.

Qi, M., T. Sun, S. Xue, W. Yang, D. Shao, and J.Mart�ınez-L�opez. 2018. Competitive ability, stresstolerance and plant interactions along stress gradi-ents. Ecology 99:848–857.

R�aberg, S., and L. Kautsky. 2007. A comparative biodi-versity study of the associated fauna of perennialfucoids and filamentous algae. Estuarine, Coastaland Shelf Sciences 73:249–258.

Remane, A. 1934. Die Brachwasserfauna. ZoologisherAnzeiger 7:34–74.

Saarinen, A., S. Salovius-Laur�en, and J. Mattila. 2018.Epifaunal community composition in five macroal-gal species – What are the consequences if somealgal species are lost? Estuarine, Coastal and ShelfSciences 207:402–413.

Silliman, B. R., M. D. Bertness, A. H. Altieri, J. N. Grif-fin, M. C. Bazterrica, F. J. Hidalgo, C. M. Crain, andM. V. Reyna. 2011. Whole-community facilitationregulates biodiversity on Patagonian rocky shores.PLOS ONE 6:e24502.

Soliveres, S., C. Smit, and F. T. Maestre. 2015. Movingforward on facilitation research: response to chang-ing environments and effects on the diversity, func-tioning and evolution of plant communities.Biological Reviews 90:297–313.

Soranno, P. A., K. S. Cheruvelil, E. G. Bissell, M. T.Bremigan, J. A. Downing, C. E. Fergus, and K. E.Webster. 2014. Cross-scale interactions:


WESTERBOM ET AL.

quantifying multi-scaled cause–effect relationshipsin macrosystems. Frontiers in Ecology and theEnvironment 12:65–73.

Tokeshi, M., and L. Romero. 1995. Filling a gap:dynamics of space occupancy on a mussel-domi-nated subtropical rocky shore. Marine EcologyProgress Series 119:167–176.

Underwood, A. J. 1997. Experiments in ecology: theirlogical design and interpretation using analysis of vari-ance. Cambridge University Press, Cambridge, UK.

Vuorinen, I., J. H€anninen, M. Rajasilta, P. Laine, J.Eklund, F. Montesino-Pouzols, F. Corona, K. Jun-ker, H. E. M. Meier, and J. W. Dippner. 2015. Sce-nario simulations of future salinity and ecologicalconsequences in the Baltic Sea and adjacent NorthSea areas – implications for environmental moni-toring. Ecological Indicators 5:196–205.

Watt, C. A., and R. A. Scrosati. 2013. Regional consis-tency of intertidal elevation as a mediator of sea-weed canopy effects on benthic species richness,diversity, and composition. Marine Ecology Pro-gress Series 491:91–99.

Westerbom, M., and S. Jattu. 2006. Effects of waveexposure on the sublittoral distribution ofblue mussels Mytilus edulis in a heterogeneousarchipelago. Marine Ecology Progress Series 306:191–200.

Westerbom, M., M. Kilpi, and O. Mustonen. 2002. Bluemussels, Mytilus edulis, at the edge of the range:population structure, growth and biomass along asalinity gradient in the north-eastern Baltic Sea.Marine Biology 140:991–999.

Westerbom, M., O. Mustonen, K. Jaatinen, M. Kilpi,and A. Norkko. 2019. Population dynamics at therange margin: implications of climate change onsublittoral blue mussels (Mytilus trossulus). Fron-tiers in Marine Science 6:292.

Westerbom, M., O. Mustonen, and M. Kilpi. 2008.Distribution of a marginal population of Mytilusedulis: responses to biotic and abiotic processes atdifferent spatial scales. Marine Biology 153:1153–1164.

Wikstr€om, S. A., and L. Kautsky. 2007. Structure anddiversity of invertebrate communities in the pres-ence and absence of canopy-forming Fucus vesiculo-sus in the Baltic Sea. Estuarine, Coastal and ShelfSciences 72:168–176.

Wood, S. N. 2006. Generalized Additive Models: anintroduction with R. Chapman and Hall/CRC, Lon-don, UK.

Wright, J. T., and P. E. Gribben. 2017. Disturbance-mediated facilitation by an intertidal ecosystemengineer. Ecology 98:2425–2436.

Yakovis, E. L., A. V. Artemieva, N. N. Shunatova, andM. A. Varfolomeeva. 2008. Multiple foundationspecies shape benthic habitat islands. Oecologia155:785–795.

Zuur, A. F. 2012. A Beginner’s guide to generalizedadditive models with R. Highland Statistics Ltd,Newburgh, UK.

Zuur, A. F., E. N. Ieno, and C. S. Elphick. 2010. A pro-tocol for data exploration to avoid common statisti-cal problems. Methods in Ecology and Evolution1:3–14.

SUPPORTING INFORMATION

Additional Supporting Information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/ecs2.2883/full


WESTERBOM ET AL.

http://onlinelibrary.wiley.com/doi/10.1002/ecs2.2883/full

http://onlinelibrary.wiley.com/doi/10.1002/ecs2.2883/full

Wave stress and biotic facilitation drive community ...

Documents

Transcript of Wave stress and biotic facilitation drive community ...