The Structure of Mediterranean Rocky Reef Ecosystems Across Environmental and Human Gradients, and Conservation Implications

OPENQ ACCESS Freely available online ~ PLOS one The Structure of Mediterranean Rocky Reef Ecosystems across Environmental and Human Gradients, and Conservation Implications

Enricsala1' ', EnricBallesteros, PanagiotisDendrinos, Antonio Di Franco, FrancescoFerretti, David Foley ', SimonettaFraschetti, Alan Friedlander, JoaquimGarrabou, Harun Guqlusoy"'"", Paolo Guidetti, BenjaminS. Halperns, Bernat Hereus, AlexandrosA. Karamanlidis, Zafer Kizilkayas", EnriqueMacpherson, LuisaMangialajos, SimoneMariani, FiorenzaMicheli, Antonio Pais", Kristin Riser", AndrewA. Rosenbergs,Marta Sales", KimberlyA. Selkoes, RickStarrs '", FionaTomas, Mikel Zabalas 1 NationalGeographic Society, Washington, D.C., United States of America,2 Centred'Estudis Avani;ats de Blanes,CEAB-CSIC, Blanes, Spain, 3 Mom/HellenicSociety for the Studyand Protectionof the MonkSeal, Athens, Greece, 4DiSTeBA, Universita del Salento,Lecce, Italy, 5 HopkinsMarine Station, Stanford University, Pacific Grove, California,United States of America,6NOAA Southwest Fisheries Science Center, Pacific Grove, California, United States of America,7 Joint Institutefor Marineand AtmosphericResearch, University of Hawaii,Honolulu, Hawaii, United States of America,8 U.S.Geological Survey, Hawaii Cooperative Fishery Research Unit and University of Hawaiiat Manoa,Honolulu, Hawaii, United States of America,9 CentreMediterrani d'Investigacions Marines i Ambientals,ICM-CSIC, Barcelona, Spain, 10 DokuzEylul University Instituteof MarineSciences and Technology, Inciralti, Izmir, Turkey, 11 SAD-EKOG,Maltepe, Ankara, Turkey, 12 NationalCenter for EcologicalAnalysis and Synthesis,University of CaliforniaSanta Barbara, Santa Barbara, California, United States of America,13 Departamentd'Ecologia, Facultat de Biologia,Universitat de Barcelona,Barcelona, Spain, 14 Universitede Nice-SophiaAntipolis, Nice, France, 18 Dipartimentodi ScienzeZootecniche, Universita di Sassari,Sassari, Italy, 16scripps Institutionof Oceanography,La Jolla,California, United States of America,17Conservation International, Arlington, Virginia, United States of America,18Estacio d'InvestigacioJaume Ferrer, IEO-Centre Oceanografic de Balears,Mao, Spain, 19 MossLanding Marine Laboratories, Moss Landing, California, United States of America

Abstract Historical exploitation of the Mediterranean Sea and the absence of rigorous baselines makes it difficult to evaluate the current health of the marine ecosystems and the efficacy of conservation actions at the ecosystem level. Here we establish the first current baseline and gradient of ecosystem structure of nearshore rocky reefs at the Mediterranean scale. We conducted underwater surveys in 14 marine protected areas and 18 open access sites across the Mediterranean, and across a 31-fold range of fish biomass from 3.8 to 118 g m !. Our data showed remarkable variation in the structure of rocky reef ecosystems. Multivariate analysis showed three alternative community states: ! large fish biomass and reefs dominated by non-canopy algae, ! lower fish biomass but abundant native algal canopies and suspension feeders, and ! low fish biomass and extensive barrens, with areas covered by turf algae. Our results suggest that the healthiest shallow rocky reef ecosystems in the Mediterranean have both large fish and algal biomass. Protection level and primary production were the only variables significantly correlated to community biomass structure. Fish biomass was significantly larger in well-enforced no-take marine reserves, but there were no significant differences between multi-use marine protected areas which allow some fishing! and open access areas at the regional scale. The gradients reported here represent a trajectory of degradation that can be used to assess the health of any similar habitat in the Mediterranean, and to evaluate the efficacy of marine protected areas.

Citation:Sala E, Ballesteros E, Dendrinos P, Di FrancoA, FerrettiF, et al.012! TheStructure of MediterraneanRocky Reef Ecosystems across Environmental and HumanGradients, and ConservationImplications. PLos ONE 7!: e32742.doi:10.1371/journal. pone.0032742 Editor:Tamara Natasha Romanuk, Dalhousie University, Canada ReceivedNovember 9, 2011;Accepted January 30, 2012; Published February 29, 2012 Thisis anopen-access article, free of all copyright,and may be freely reproduced, distributed, transmitted, modified, built upon,or otherwiseused by anyonefor any lawfulpurpose. The work is madeavailable under the CreativeCommons CCO public domain dedication. Funding:This study was funded by the OakFoundation, the PewCharitable Trusts, Spain's National Council for ScientificResearch CSIC! and the LenfestOcean Program.The funders had no role in studydesign, data collection and analysis,decision to publish,or preparationof the manuscript. CompetingInterests: The authors have declared that no competinginterests exist. ' E-mail:esalalings.org

Introduction because there is no rigorous historical baseline for the abundance of marine species or the suucture of marine ecosystems in the Intense exploitation over millennia has depleted Mediterranean Mediterranean [6,7], except for 8 few taxa and local time series of species from the large to the small, including the Mediterranean fishery dependent and independent data P]. Most of the monk seal, sea turtles, bluefin tuna, groupers, red coral, lobsters, quantitative data on the structure of Mediterranean ecosystems and limpets e.g., [1,2,3]!. Habitat destruction, pollution, intro- originates from field studies in the last 30 years. Therefore, our duced species and climate change have also taken 8 toll on attempts to evaluate the health of the marine ecosystem and the Mediterranean species and ecosystems [4,5]. Although these eAicacy of recent conservation actions at the ecosystem level are impacts have been significant, based on qualitative observations constrained by 8 limited sense of what is possible or natural [8]. over the millennia, it is diAicult to evaluate their magnitude Here we establish the first current comparable baseline of

'... PLoSONE ~ vvvvvv.plosone.org February2012 ~ Volume 7 ~ Issue 2 ~ e32742 Mediterranean Rocky Reef Baselines ecosystem structure at the Mediterranean scale, focusing on frequent, faunas are shifting, and invasive species are spreading nearshore rocky reefs. P2,33]. What would a 'healthy' Mediterranean rocky bottom look like? There has been a great deal of work on the direct and indirect There are no pristine sites i.e. undisturbed by humans, with effects of human impacts on fish and benthic communities in the historical ecosystem structure and carrying capacity! left in the Western Mediterranean. However, most studies have investigated Mediterranean that allow us to set a baseline against which to the individual effects of a number of natural and human compare the health of current ecosystems. Research on pristine, disturbances, and processes driving observed changes remain historically unfished sites in the central Pacific show that intact, untested across appropriately large ecosystem and geographical complex reef ecosystems harbor large biomass of fishes, with scales P4]. Experimental manipulation of fishing, pollution, inverted biomass pyramids, and high coral cover [9,10]. Fishing habitat degradation, invasions and climate change impacts across pressure has been a major stressor on Mediterranean reef systems. the Mediterranean is not feasible. The only practical way to Thus, in the Mediterranean, we would expect total fish biomass to address the spatial distribution of community structure and its be also the single most important indicator of the health of fish relationship with environmental and human factors is by populations, with biomass increasing with decreased fishing measuring community structure across gradients of these factors. pressure, as Mediterranean no-take marine reserves demonstrate Examples of successfulrecovery at the ecosystem level are rare [11,12,13,14]. Therefore, marine reserves are the best proxies for for the Mediterranean and are systematically related to the the trajectory of recovery of fish assemblagestowards a pristine presence of marine protected areas MPAs! [15,35]. This study state, possibly including cascading effects leading to a wider provides the first current baseline and trajectory of degradation recovery of the protected ecosystems. However, we expect these and recovery for Mediterranean rocky reefs, against which the current baselines to be still far from historical baselines with an present condition of existing and candidate marine protected areas intact ecosystem likely including all apex predators such as sharks can be assessed.More broadly, we provide a framework and and monk seals. methodology for establishing such gradients of degradation in the Predatory fishes can have a major role in determining the absence of historical data for the Mediterranean and other marine abundance of their prey and strongly modifying the ecosystem. In ecoregions. This Mediterranean-wide baseline should allow us to the Mediterranean, these effects have been observed on sea understand the magnitude of human impacts on nearshore rocky urchins, which are the major benthic herbivores on Mediterranean reefs, a key habitat with critical functional importance, and better rocky bottoms [15,16]. At high predatory fish abundance, evaluate the results of management and conservation actions. predation tends to maintain low sea urchin abundances, while at low predatory fish abundance, sea urchin abundance is regulated Methods by many other factors and thus their abundance becomes less predictable [12,17]. The Mediterranean has only two major native Habitat, sampling sites and spatial replication herbivorous fishes, Sarpasalpa and Sparisomaoretense [IB]. Although We conducted SCUBA surveys in May June 2007 and 2000 at at large abundances Sarpa salpa should be able to reduce the 13 MPAs and 17 unprotected areas across the Mediterranean biomass of some benthic algae [19,20], only introduced herbiv- Spain, Italy, Greece, Turkey; Fig. 1, Table SI!. One additional orous fishes Siganusspp.! have been shown to cause strong algal MPA and one fished site were surveyed in Morocco in April 2010, declines to the extent of creating barrens! in the Eastern for a total of 32 sites. We sampled rocky habitats at 0 12 m depth, Mediterranean [21]. The decrease of these algal communities at 4 6 replicate stations within each site, according to their can also affect the recruitment rate of numerous rocky fishes that extension. select algae as settlement habitats [14], having a potential cascading effect on the whole community. We would then expect Fish censuses a complex, near pristine benthic community with low abundance Fish data were collected using standard underwater visual of sea urchins and large algal biomass. census techniques e.g., [9]!. Sampling stations within sites were Mediterranean shallow benthic communities harbor hundreds spaced at least 1 km apart from the next, except in very small of species of algae and invertebrates, but they tend to be marine reserves e.g., Portofino! where we sampled stations dominated in cover and biomass by algae [22]. In particular, the hundreds of meters away in order to have enough replicate least impacted communities are often dominated by canopies of surveys. We conducted three replicate 25 m-long and 5 m-wide Fucales, mostly C~stoseiraspp. [4,23,24]. The abundance of transects at each station. Along each transect, the diver swam one C~stoseiraappears to be determined by multiple factors, including way at constant speed, identifying and recording the number and water quality, sea urchin grazing, coastal development, and size of each fish encountered. Fish sizes were estimated visually in historical and current fishing pressure [25,26,27,20,29]. Fucales 2 cm increments of total length TL!. Fish biomass wet mass! was suffered a long-term decline in the NW Mediterranean in the last estimated from size data by means of length-weight relationships century due to a combination of the above direct and indirect from the available literature and existing databases P6,37]. We human impacts [26]. Introduced algae have been present in the focused on fish biomass in our analysis instead of fish density Mediterranean since the nineteenth century, but their number and because biomass is the single most important indicator of the impact on native benthic communities has increased exponentially health of fish assemblagesacross gradients of human disturbance over time P0,31]. e.g., [9,30]!. For our analysis, we assigned each fish taxon to one In addition to the direct and indirect impacts of overexploi- of four trophic groups using the information about diet in the tation, there have been other major impacts to Mediterranean literature P7], and in previous Mediterranean studies [12,39]: nearshore reefs. Historically, land use changes in the Mediter- apex predators, carnivores, herbivores and detritivores, and ranean region had accompanying changes in nutrients and planktivores Table S2!. sedimentation, and a major loss of coastal habitats [4]. The Mediterranean is also increasingly affected by climate change. Benthic algae Seawater temperatures are steadily increasing, extreme climatic We scraped all non-encrusting algae in five replicate 20 x20 cm events and related disease outbreaks are becoming more quadrats at each station. Each sample quadrat! was placed in an

'... PLoSONE ~ www.plosone.org February2012 Volume 7 Issue 2 e32742 Mediterranean Rocky Reef Baselines

NorthAtlantic Ocean

BlackSea GEN~ POR CRE % MON TRE 43'14 MEO TGC PORCAV CAP eTAV oRAI ~ ~ ~ PIP EIV~ AIR PCBOTR ~ e Avv PIHN~ CAB CAR MAR AI.O PET WVA GOK~ ~ AI-PALH KAB 35'N KAI-I MediterraneanSea

~ Unprotectedsites ~ Marineprotected areas

5'E 10'E 15'E 23'E 25'E 30'E

Figure 1. Map of study locations. ADR=Adrasan,AIR= illa de I'Aire, ALH=Al-Hoceima unprotected!, ALO=Alonissos,ALP = Al-HoceimaMPA, AYV= Ayvalik, CAB= Cabrera,CAP = Capo Caccia,CAR = Carloforte, CAV= Cap de Cavalleria,CRE = Cap de Creus, DRA= Dragonera, EIV= Eivissa, FET= Fethiye, FMN= Formentera-Espardell, FOR= Cap Formentor, GEN= Genoa, GOK= Gokova, GYA= Gyaros, KAR= Karpathos, KAS= Kas, KIM= Kimolos-Polyaigos,MAR= Maratea, MED=Medes Islands, MON=Montgri, OTR=Otranto, PCS=Porto Cesareo,PIP= Piperi, POR=Portofino, TAV= Tavolara,TGC= Torre Guaceto,TRE= Tremiti. For details on protection level see Figs.2 3 and Table l. doi:10.1371/journal.pone.0032742.g001 individual zip-lock bag and brought to our field laboratory. After the data on area covered by benthic groups to wet mass using erect algae were removed from a quadrat underwater, the diver conversions from the literature [41]. For analysis we pooled taxa estimated the cover as % of the quadrat! of encrusting algae. After into larger functional groups see Data Analysis!. the dive, sampleswere stored in a cooler and transported to the field lab in seawater, without any preservative. Algal biomass was Environmental and human-related data measured the same day of collection. Samples were individually We collected habitat and environmental data for all stations in sorted and weighed. Water excessfrom every sample was removed situ, including latitude, longitude, bottom rugosity, degree of using a salad spinner; sampleswere spun for 30 secondseach. Erect exposure, and depth to try to minimize variability in the benthic algae were sorted and weighed to species on a digital balance and fish community due to habitat alone [11]. ~0.1 g!. Turf algae were identified to speciesin the field lab when We measured the rugosity of the bottom or surface relief! as an possible, otherwise weighed as a group. Introduced species were indicator of the structure of the physical habitat, which is known to always identified to species and weighed individually. We have an important role in determining the abundance of fish e.g., considered all introduced species mentioned in the scientific [II,43]!. To measure rugosity, a 10-m long small link chain literature [40]. The biomass of encrusting algae was determined .3 cm per link! was draped along the length of the centerline of using the cover estimated in the field and conversion rates from the each transect [44]. Care was taken to ensure that the chain literature [41]. Algal biomass was pooled into the following groups followed the contour of all natural fixed surfaces directly below the for data analysis: native! canopy algae Fucales: C~stoseiraspp. and transect centerline. A ratio of 10 to the linear horizontal distance Sargassumspp.!, erect algae e.g., Dictyotales!, turf algae e.g., small between the beginning of the transect and the end of the draped filamentous algae!, encrusting corallines e.g., Lithophpllum!,encrust- chain gave an index of rugosity. ing non-corallines e.g., Lobophora!,and introduced algae. To measure the influence of the oceanographic climate we obtained 8-year averages of satellite-derived mean and median Benthic invertebrates seawater surface temperature SST! and net primary productivity Sea urchins: we recorded sea urchin density and size I cm size PP!. SST was derived from the Pathfinder Version 5 Advanced classes, test without spines! using a plastic caliper on thirty Very High Resolution Radiometer AVHRR! data set [45]. 50 cmx50 cm quadrats per station. Biomass was then estimated Vertically integrated PP was calculated by applying the method of using size-biomass conversion rates from the literature [42]. Behrenfeld and Falkowski [46] to surface chlorophyll-a concen- Sessile invertebrates: we estimated the cover of major taxa of tration, photosynthetically available radiation and sea surface sessileinvertebrates using the point-count method on a 25 m-long temperature provided by the Moderate-resolution Imaging transect with marks every 20 cm. We identified the organisms to Spectrometer MODIS! carried aboard NASA's Aqua spacecraft. the lowest possible taxonomic level. For instance, most sponges Time series data from July 2002 June 2010 were extracted from and tunicates were identified to species, although smaller colonial monthly averagesusing 5 x5 km boxes centered on each study site. organisms such as bryozoans were identified as a group when Means and medians for each site were then produced from these identification in situ was not possible. We then transformed the time series.

'... PLoSONE l www.plosone.org February2012 l Volume 7 l Issue 2 l e32742 Mediterranean Rocky Reef Baselines

Table 1. Characteristics of the marine protected areas investigated in this study

Total MPA Sizeof no-take Year of Age of reserveat Enforcement Name Zone sampled size ha! area ha! creation samphng years! level

Alonissos-Northern Partialprotection 207000 438 1992 16 Low sporades Cahfieta:...:::'.:: .:::::::'.:::::::::::::: .::::::.:::::::::::Pi al ial:::prytefc t ion::::::::::::::::: .:::::::::'.1002%::::::: .::::::::.::::::::.::::::::::::::36tt::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::E929 1 Nordde Menorca Partialprotection 5119 1100 1999 8 High Cavalleria! C'a'p'.d'e'.:'C're'u's.::'::.:':.:.:':.:::.::.:': P'a'r'1 'pipe'c'tio'r'f.:': 'al': ':.::.:'.:: 308'6.::.:' '.:'.:'.:' ' '.:'.:''.:'.:'.::'.:.:.:' ' '2k' ':'::::::::::::::::::::::::: 1'998.::.::::::::::::::::::::::: 9 CapoCaccia No-take 2631 38 2002 Medium Fp'ime'nteir''a.'-:'E's'p'a'rd'el:N'o-take .::.::.::.::.:.:'. t'3'6'17.:'.:::.::.::::.::''::.:''.::'::.:400: :'.::::: : :'::.:'.:'.:'.:'.:'.:'::.:'.:''' :'.:''::.::.::'1'999::.:':.:.::. :':.:':.:' '.:'.:'.:' '.:''8:.::'.::'.:'.:'.:'.:' '.:'.:'.:'' '.:'.::'.:'.:'.:'.:'.:'.:'.:'.::'.::'.::'.::'.::'.::'.::'.::'.:'.::'.::'.::'.::'.::'.::'.::'.::'.:.:.:. :Hi'gh MedesIslands No-take 94 94 1983 24 High I?superi::.::::::,.::::::,.:,::,.:,:::,.::,::,:::,:.:,::::,:::.:,:.:::,::,::::,:::::::,::::::::,No:-:take::,:::::.:,::,:.:,::,::,:.:,:.:,:.:,:.:,::,:.:,:.:,:::.:,.::::,::,::.::,::,.::,.::,.::,::,::,438,::.:.::.::,.::.::.::.::,.::.::.::.::,.::.::.::.::.::.::.::.::.::.::.::.::.:::438 PortoCesareo No-take 16654 173 1997 10 Low Higth Tavolara No-take 15357 529 1997 High Torrid:.:6 facetot::::::::::::::::::::::::::::::::::,:::::::::.::::Neptune.::.::.::.::.::.::.::.::.::.::::.::.::.::::.:::::::::::::::.:::::::::::227::::::::::::::::::::::::::::::::::::::::::::::.:TiHigth 9 Tremiti No-take 1466 180 1989 19 Medium

All thesemarine protected areas had partial protection, with areaswhere some types of fishingare permitted, and small no-take areas. Piperi is the no-takearea of the MarinePark of Alonissos-Northernsporades. Enforcement level sensu Cuidetti et al. 8]. doi:10.1371/journal.pone.0032742.t001

To measure combined human impacts at the study sites we used but here we report only the level of protection at the time of our a cumulative human impact score determined via an assessmentof sampling. Unprotected sites are typically open-access with little the cumulative impact to Mediterranean marine ecosystems enforcement of fishing regulations. To minimize differences resulting from multiple pressures, including fishing, climate possibly deriving from other human threats combined to fishing, change, pollution and biological invasions. This analysis builds sites were selected within areas not directly affected by evident on a previous global analysis of cumulative human impacts that sources of impact e.g. harbors, defense structures, sewages,strong involved combining global spatial data on 17 types of human urbanization!. pressures, the distribution of 20 marine ecosystem types, and scores representing the potential impact of each pressure on each Data a nalysis ecosystem type derived through an expert judgment survey Univariate analysis on total fish biomass was conducted to approach [47]. Impact scores determined for each pressure- examine differences among sites and stations, and to test whether ecosystem type combination are summed, within each 1 km of fish biomass responded to protection. Effects of protection were the ocean, to calculate the cumulative impact score. Micheli et al. analyzed using three-way permutational analysis of variance in prep.! replaced some of the data layers and included additional PERMANOVAl [49] on square root-transformed data and based data to better reflect the specific pressures and ecosystems of the on Euclidean distance dissimilarity matrix. The sampling design Mediterranean basin. A total of 22 spatial datasets of human consisted of 3 factors: protection fixedl, site nested in protection!, activities and stressors and 19 ecosystem types were assembled and station nested in site!. For the PERMANOVA we used three and used in the analyses http: //globalmarine.nceas.ucsb.edu/ levels of protection: no-take reserves, MPAs that allow some mediterranean/!. Here, we used the footprint i.e., the cumulative fishing, and unprotected areas. Finally, fish taxa were pooled into impact score! of all pressures acting on a 3-km radius around each trophic groups Table S2l because fishing disproportionately of our field sites as a measure of cumulative human impact on our targets species at higher trophic levels [50], and recovery from focal rocky reef ecosystems. See Supporting Information Sl for fishing potentially includes increased abundances or biomass of more details on the cumulative human impact analysis. high-level predators and shifts in trophic structure [51]. To The MPAs studied here include a range of protection levels investigate the relationship between fish biomass in MPAs and age from no-take marine reserves to areas with virtually no of the MPA, size of the reserve, the level of protection and regulations! and enforcement levels Table Il. Based on available enforcement, and environmental variables we used multiple linear scientific information, personal experience and knowledge of the regression analysis [52]. MPAs, and interviews with MPA staff reporting on the overall To test for differences in biomass of algae between sites we used management effectiveness [40], MPAs were categorized as follows: a non-parametric Kruskal-Wallis median test. To investigate the al well-enforced no-take reserves Formentera-Espardell, Medes, relationship between pairs of trophic groups, we used linear and Portofino, Torre Guaceto, Tavolaral, bl MPAs where some generalized linear models and their extensions to model linear and fishing is allowed or some fishing occurs due to weak enforcement non-linear relationships, complex trajectories, heteroscedastic data Cabrera, Cap de Creus, Capo Caccia, Porto Cesareo, Cavallerial, and a large inter-site variation, which we incorporated by and cl poorly-enforced MPAs AI-Hoceima, Alonissos, Piperi, assuming that the site effect was a random variable following a Tremitil and open accessareas Table Il. Some of our study sites normal distribution N, o l. Although in most casesvariance of may have been protected after our study period e.g., Karpathosl, the different variables increased with density, log-transforming the

'... PLoSONE l www.plosone.org February2012 l Volume 7 l Issue 2 l e32742 Mediterranean Rocky Reef Baselines variables usually stabilized the residuals and linearized the relationships. When transformation did not linearize the relationships, we assumed a Gamma distribution for the regressed variable, i.e. we assumed that the sampled value of the regressor was a draw from a random variable with mean lt and variance lt2/ v, where v is a dispersion parameter. We also tested the presence of complex relationships between the variables by using Gener- alized Additive Mixed Models GAMMj. By doing so we had a certain degree of freedom in testing correlations between variables, and the opportunity to detect the shape of significant relationships. To describe the pattern of variation in community structure patterns of distribution of biomass/abundance of functional groups within the community! and their relationship to environmental and human gradients we used linear ordination methods. Linear models are appropriate for our data because a preliminary detrended correspondence analysis showed short gradient lengths -1.0 1.0 <2 SD! [53]. To explore the spatial distribution of community structure across the Mediterranean and its relationship with environmental variables we performed a direct gradient analysis redundancy analysis: RDA! [54] on log-transformed data using the ordination program CANOCO for Windows version 4.0 [54], and the species and the environmental data matrices. The RDA introduces a series of explanatory environmental! variables and resembles the model of multivariate multiple regression, allowing us to determine what linear combinations of these variables determine the gradients. We pooled data from all taxa into the following groups to facilitate the large-scale analysis: biomass of fish trophic groups, canopy algae, other algae, introduced algae, sea urchins, suspension/filter feeders, and cover of bare rock. The environmental data matrix included the following variables: mean surface seawater temperature, mean primary productivity, bottom rugosity, protection level, and cumulative human impact. Latitude was significantly correlated to mean and median SST Spearman rank order correlation, p< 0.001, r = 0.56 and 0.48, respectively! and mean and median PP p<0.01, r = 0.24 and 0.22, respectively!, and longitude was -1.0 1.0 significantly correlated to mean and median SST p<0.001, r = 0.56 and 0.40, respectively!, thus we did not use latitude and Figure 2. Biplot of results of redundancy analysis on biological and environmental data. Site codes as in Fig. l. Green circles are longitude in the analysis. To rank environmental variables in their well enforced no-take reserves,yellow circles are marine protected importance for being associatedwith the structure of communities, areaswhere somefishing is allowed or some fishing occursdue to weak we used a forward selection where the statistical significance of enforcement, red circles are non-enforced MPAs,and white circles are each variable was judged by a Monte-Carlo permutation test [55]. unprotected, open accessareas. doi:10.1371/journakpone.0032742.g002 Results the more important groups were apex predators 81%! and Our data reveal three groups of sites mainly constituted by: I! carnivores 5%!, and bare rock 0%! in the opposite direction. well-enforced no-take marine reserves with high fish biomass, ! Only two variables were significantly correlated to the RDA partial marine protected areas and weakly enforced no-take axes Table 3!. Primary production PP! was significantly marine reserves with lower fish biomass, and ! non-enforced correlated with the first axis, indicating a negative relationship marine protected areas and areas open to fishing Fig. 2!. The between PP and abundance of canopy algae and suspension main factor involved in this ordination is the protection level, feeders. PP did not have a significant relationship with the which is largely correlated with the biomass of apex predators and abundance of invasive species, or with the extent of sea urchin carnivores Figure 2!. Our data also revealed a gradient along barrens urchins and bare rock; Fig. 2!. Protection level was three alternative ecosystem states: I! ecosystemscharacterized by significantly correlated with the second axis, indicating that less large fish biomass and benthic communities dominated by non- fishing is associated with larger fish biomass especially with apex canopy algae ' predator-dominated' ecosystems, i.e., the well- predators and carnivores!, and more fishing is associated with less enforced marine reserves!; ! ecosystemswith lower fish biomass, fish, more sea urchins although weakly! and bare rock. In but abundant algal canopies and suspension feeders; and ! contrast, the abundance of canopy algae was independent of the ecosystems with lower fish biomass and extensive barrens, and level of protection. areas covered by turf algae Fig. 3!. Total fish biomass was significantly different among sites and The first two axes of the RDA biplot explained 97% of the stations PERMANOVA test, p<0.001 for both factors!, and species-environment relationship Fig. 2, Table 2!. In terms of ranged between 3.8 and 118 g m, a 31-fold difference Fig. 2!. community structure, the fit of canopy algae Fucales! in the first Total fish biomass was significantly greater in no-take marine axis of the RDA ordination diagram was 31%; in the second axis, reserves than in MPAs that allow some fishing p<0.001!. The five

'... PkoSONE ~ www.plosone.org February2012 ~ Volume 7 ~ Issue 2 ~ e32742 Mediterranean Rocky Reef Baselines

Figure 3. Examples of the major community states identified in this study: large fish biomass and non-canopy algae 'predator- dominated' ecosystems, i.e., the well-enforced marine reserves A!; ecosystems with lower fish biomass, but abundant algal canopies and suspension feeders B!; and ecosystems with lower fish biomass and extensive barrens C!, and areas covered by turf algae D!. Photos:A,B: E Sala,C: Z Kizilkaya,D: A Verges!. doi:10.1371/journal.pone.0032742.g003 highest ranking sites, in terms of fish biomass, were the well- sharks in any of our study sites, although we observed individual enforced no-take reserves Tavolara, Medes, Portofino, Torre monk seals /Ifonachus monachus!at Kimolos-Polyaigos, Adrasan, Guaceto and Formentera-Espardell! Fig. 2!. In contrast, fish and Karpathos. biomass in MPAs that allow some fishing were not significantly The biomass of benthic algae was significantly different among different than unprotected sites PERMANO't/A pairwise test, sites Kruskal Wallis median test, p<0.001! and ranged between p=016! Fig 2! 477 g m at Fethiye and 4565 g m at Carloforte, a 10-fold The biomass of every fish trophic group was correlated with difference Fig. 4!. In contrast with fish biomass, there was no total fish biomass Spearman rank order correlation on sampling apparent relationship between benthic community biomass and station averages, p<0.0001!. Total fish biomass was positively composition and the protection status of sites. Some unprotected related to level of protection p = 0.030! but not to age and size of sites characterized by extremely low fish biomass had high algal the reserve or any other variable multiple regression, R = 0.70, biomass and well developed algal canopies e.g., Maratea and p = 0.16! Table 4!. The proportion of total biomass accounted for Kimolos!, whereas the well-enforced reserves with the exception by apex predators ranged between 0'/o at Tremiti and 49'/o at the of Formentera-Espardell! had generally low abundance of canopy- Medes Islands Marine Reserve Fig. 4!, and the biomass of apex forming algae Fig. 4!. predators increased nonlinearly with increasing total fish biomass The relationship between total biomass of algae and biomass of R = 0.70, p<0.001! Fig. 5!. Apex predator biomass, on average, canopy algae Fucales! shows two different patterns, with some was significantly larger in protected sites than at unprotected sites stations showing a non-linear increase after 2000 g m of total 0.4'/o vs. 5.5'/o, chi-square test, p<0.001!. We did not observe algae, and other sites remaining at zero biomass of Fucales Fig. 5!. The latter are sites located in the NW Mediterranean Medes, Montgri, and Cap de Creus!. For sites where Fucales are present Table 2. Results of redundancy analysis RDA! on log- there was a log-linear relationship between total algae and Fucales transformed data on fish biomass, benthic biomass and cover of bare rock, and environmental and human-related variables. Table 3. Conditional effects of the Monte-Carlo permutation results on the redundancy analysis RDA!. Total Axes 1 2 3 variance

Eigenvalues 0.139 0.091 0.007 1.0 Variable LambdaA F P Species-environmentcorrelations 0.626 0.738 0.241 Protectionlevel 0.10 3.18 0.006 Cumulativepercentage variance MeanPP 0.09 3.41 0.010 of speciesdata 13.9 23.0 23.7 Humanimpact 0.05 1.68 0.094 of species-environmentrelation 58.7 97.0 100 MeanSST 0.04 1.55 0.178 Sumof all eigenvalues 1.0 Rugosity 0.02 1.01 0.408 Sumof all canonicaleigenvalues 0.237 PP:primary production, SST: surface seawater temperature. doi:10.1371/journal.pone.0032742.t002 doi:10.1371/journal.pone.0032742.t003

'... PLoSONE ~ www.plosone.org February2012 ~ Volume 7 ~ Issue 2 ~ e32742 Mediterranean Rocky Reef Baselines

Table 4. Results of multiple linear regression on total fish even lower than the most overfished coral reefs in the Caribbean biomass in the studied marine protected areas. Pfl]. Apex predators are an important component of the fish assemblage by biomass at the sites with largest fish biomasses, Variable Partialcorrelation p reaching a maximum of 49% of total fish biomass at the Medes Islands Marine Reserve. This is similar to the structure found at the Papahanaumokuakea Marine National Monument in the NW Reserveage 0.28 0.323 Hawaiian Islands [9], and approaching the inverted biomass M'e''an::5:5T: pyramid found at unfished reefs in the central Pacific [10,57]. MeanPP 0.06 0.926 Moreover, biomass of apex predators such as the dusky grouper Humaii:.:impact:outsi'd'e.. :'iese'rv'e.:::.::,:::.::.:.:,::,::.:.::,::,:.:.: '.:,':,'0:92.::.:..::.::.::.:..::.::.::.::.::.::.::.::.::.::.::.::.::.::.::.:..:..:..:..:..:..:..:..:..:..:0 95'4 Epinephelusmarg~natus! continues to increase at the Medes Island Marine Reserve after 27 years of protection [50]. These results Levelof protection 0.90 0.038 suggestthat well enforced Mediterranean marine reserves are in a Bottom::iu''g'o'sity'. trajectory of recovery similar to reserves elsewhere that moved from a degraded state with low fish biomass to biomass values doi:10.1371/journal.pone.0032742.t004 similar to unfished sites e.g., [59]!. As we expected, the highest levels of fish biomass were found slope = 0.00023, random effect o = 0.545, p<0.0001!. only inside well-enforced no-take marine reserves. An important Suspension feeders had very low biomass across sites compared result of our study is that MPAs where fishing is allowed and to benthic algae, from 0.7 g m to 31 g m for a complete list weakly enforced no-take marine reserves had fish biomass of benthic taxa encountered during the surveys see Appendix 2!. comparable to unprotected sites. The worst performing MPA The average percentage of the substrate with bare rock ranged was Morocco's Al-Hoceima, which is a 'paper park' with virtually from 0% substrate totally covered by benthic organisms! to 31%. no management and enforcement see [40]!. Al-Hoceima was the The sites with largest cover of bare rock were Fethiye 1%!, Kas westernmost site in our study, and it is subject to influence of 0%!, and Gokova 7%! ANO't/A, p<0.001!, which harbored Atlantic waters, which makes it the most productive of all our sites, barrens caused by Siganidae [21]. making this result even more striking. Sea urchin biomass also differed significantly among sites MPAs were located throughout the multivariate space, although Kruskal Wallis median test, p<0.001!, ranging from 0 to they were clearly differentiated by their effective level of 533 g m Fig. 2!. There was no significant relationship between protection. As indicated by Guidetti R. Sala [12], MPAs may the biomass of carnivorous fishes or the biomass of the major have larger fish biomass than adjacent unprotected areas, but not predators Diplodus spp. [56]! and sea urchin biomass, even when necessarily larger biomass than unprotected areas located accounting for the site effect and the non-normality of the data elsewhere. This is because of the history of fishing in each locale, using GLMM p-value = 0.02!. Presence-absence of urchins was local productivity, the effectiveness of MPA enforcement, and not related to carnivores either. Positive occurrences of urchins other factors that affect the overall local potential to support fish were not related to carnivores, protection level, and the cumulative biomass. Therefore, MPAs and unprotected areas in the human impact score Supporting information Sl!. There was a Mediterranean should not be regarded as two opposite situations; substantial variability between sites .165 times the within-site instead, each site should be considered within a continuum from variability!. most degraded to healthier regardless of whether it is protected. There was no significant relationship between the biomass of Otherwise, the mere existence of partially protected MPAs may benthic communities and total fish biomass, and between give the false impression that conservation is occurring. herbivorous fishes and total algal biomass at the regional scale The fish biomass levels measured in this study are not the largest Supporting information Sl!, although in the Eastern Mediterra- in the Mediterranean e.g., [I I,I3]! although they may harbor nean, in particular in Turkey, biomass of herbivorous fishes was larger biomass values following seasonal variations. Some sites, negatively correlated with algal biomass [21]. Unexpectedly, there such as the Medes Islands, have significantly larger biomass in was also no significant relationship between sea urchin and total August-September when warmer waters bring piscivorous fishes to algal biomass Figure S 1!. shallow waters e.g., Dentex den ex,Seriola dumerilr! unpublished data!. Therefore the ranking of fish biomass among the well- Discussion enforced marine reserves calculated in this study may exhibit some changes due to seasonal changes in fish biomass. The largest Fish assemblages published biomass, at the Cabo de Palos Marine Reserve Our results reveal remarkable variation in the biomass and 39 g m [11]! and other reserves,is restricted to very particular structure of coastal ecosystems and provide the first empirical habitats offshore with significant water motion and currents, and trajectory from degraded to healthy states in the Mediterranean. probably not typical of most Mediterranean rocky reefs. These Such a trajectory is most clearly associated with variation in total reserves have recovered successfullyrelative to nearby unprotected fish biomass. We found a 31-fold range in fish biomass on rocky sites, and their biomass is larger than the best preserved reserves in bottoms across the Mediterranean. If we consider the largest fish Kenya [60], similar to the best preserved sites in the Caribbean biomass values estimated in the Mediterranean at comparable Pfl], but still below the best available baselines for coral reefs in habitats and depth 39 g m [11]!, the gradient of fish biomass the central Pacific, where unfished islands have biomass levels up is an extraordinary 115-fold. This is the largest fish biomass to 000 g m [10]. gradient ever reported for reef fish assemblages e.g., [10,30]!, and Marine reserves with the largest fish biomass are a useful is indicative of the large impact of historical and current intense current baseline against which managers can compare recovery fishing pressure in the Mediterranean. The lowest biomass values trends for fish assemblages in rocky habitats across the in Gokova Bay, Turkey, and Al-Hoceima, Morocco! are among Mediterranean, which can be further refined by data coming the lowest reported in the scientific literature for shallow reefs, from past and future studies able to identify the main drivers of

'... PLoSONE ~ www.plosone.org February2012 ~ Volume 7 ~ Issue 2 ~ e32742 Mediterranean Rocky Reef Baselines

@ Well-enforced~p,p no-take reserves 100 I- Other "MPAs" 80 E 0 HE 60

R 40 5 o 20 I I ~! . ~I ~I I ~ I~IIIIII ~ Q IZ V Z I'-L Lu IZ: CCCCI rr E Z > CL Lu Z > < > P i cL CCg Vl CZ:I CCcL Z + Lu O c0 Z < tZ < I- < O < p ~ cL CCLu >- CC< » < O < O Lu < ~ ~ p ~QI-~VVOQV~Y% iQ<00<0

5000 ~ INTRQOUCED ENCRUSTING TURF ' ERECT ~ CANOPY

CV 4000

3000

E202000 otal rs Q ~O 1000

0 ! 2 X U Z CLLu CC 2 < > X Z ! > LLIZ ! < ! O CX Xz IrIX + 2 CLI O O O Q 2 0 2 I '< < CI '' < 3 g + Y ~ u-

600-

500 E0400

J3 300 C:

LJ 200

rtI CLI 100

Q 0' 0 Z 0 LLI CCCZ: CQ S tt. 'Z ! 0 LLI Z > < > 0 4 L' CZ:S 0! CL'I- 0' 0- Z Y «CI C02 < Iz. < I- < o < o ~ 0 tt. Lu >- tz. < ~ >- < CI < o Lu <» o 0 I u UO C2OOCL Y g I-C0 < OO

Figure 4. Biomass of major taxa investigated in this study. AP=apex predators, CA= carnivores,ZP= zoojplanktivores, HE= herbivores+ detritivores. doi:10.1371/journal.pone.0032742.g004

'... PLoSONE ~ www.plosone.org February2012 ~ Volume 7 ~ Issue 2 ~ e32742 Mediterranean Rocky Reef Baselines

1,6

I.2 X3 Q G.8 "a

Q, GA

I,G

i;GQ;, 4 tri

HOGG

P3 «QGG -I

c 2GGG -;

G lUGQ 2GQG 3QrJQ 4GQG 5QGG 69GQ iQGG Total algal biomass g/m2!

Figure 5. Top: Relationship between total fish biomass and biomass of apex predators. Bottom: Relationshipbetween total biomassof algae and biomass of canopy algae Fucales!.Sites with no Fucalesabove 2000 g m are found in the NW Mediterranean. doi:10.1371/journal.pone.0032742.g005 recovery and degradation. It is clearly the case for the Western em to the southeastern Mediterranean, our multivariate analysis Mediterranean, where fish biomass achieved the highest values in showed that PP was not a significant factor in determining fish the dataset but also the lowest, at AI-Hoceimal, which is consistent biomass across the Mediterranean only the level of protection with documentation of successfulrecovery in no-take reservesfrom was. Moreover, PP values at Piperi and other Eastern Mediter- this ecoregion e.g., [12,13,61]l. However, there is no good ranean sites are similar to the Balearic Islands, where well- estimate of the maximum potential fish biomass in the Eastern enforced no-take reserves e.g., Formentera-Espardelll have Mediterranean. Piperi, the only no-take marine reserve as the recovered fish biomass to values twice as high as Piperi. "core area" of the National Marine Park of Alonissos, Northern Conversely, PP is high at Al-Hoceima, Morocco, which had the Sporadesl in the Aegean, has low values of fish biomass, which are lowest fish biomass recorded. In any case, because of the absence comparable to unprotected areas. Piperi has dismal enforcement, of well-enforced no-take reserves in the Southern and Eastern and during our field surveys we observed many fishing lines Mediterranean it is currently not possible to analyze the role of tangled on the bottom as well as fishing spears stuck on rocks. productivity in determining the recovery of marine reserves and Although primary productivity PPl declines from the northwest- their maximum biomass in the Mediterranean.

'... PLoSONE l www.plosone.org February2012 l Volume 7 l Issue 2 l e32742 Mediterranean Rocky Reef Baselines

Habitat structure has been shown as an important factor in C~stoseiracanopy after protection comes from the Medes Islands determining the structure of fish assemblagesin the Mediterranean Marine Reserve. The Medes Islands did not have sublittoral e.g., [11]!. However, we found that rugosity was not positively Cjstoseirawhen they were protected in 1983, but Cjstoseirasp. correlated with total fish biomass or with the structure of the whole became abundant after 1992 [67,68], suggesting that recovery of community fish+benthos! at the Mediterranean scale. These formerly abundant C~stoseiracanopies in the NW Mediterranean results indicate that the level of protection, not local habitat [26] takes longer than recovery of fish assemblages. Since characteristics, is the most important factor in determining total dispersion of C~stoseiraappears to be very limited, the recovery fish biomass in the Mediterranean. of lost canopies in large areas may prove difficult [69]. Our results are restricted to a shallow depth range 8 12 m!, The Scandola Natural Reserve in Corsica, which we did not and comparisons with other Mediterranean locales should be survey in this study, has a no-take area with large fish biomass [70] restricted to similar depths. We observed larger fish biomass a little and dominant C~stoseiracanopies [71]. Scandola is probably the deeper in some of the study sites. For example, large dusky best example of a 'healthier' rocky reef, without fishing and with groupers Epinephelusmargi'natus! were more abundant at 15 20 m good water quality, which made it one of the Mediterranean's depth in the Medes Islands Marine Reserve and Cabrera National 'reference' sites [25]. Nevertheless, we do not know whether Park; the goldblotch grouper E. eostaewas rarely observed in our recovery of C~stoseiraat sites where it disappeared historically is transects in Turkey, but we commonly observed adults below facilitated by increased fish biomass within marine reserves. The 25 m depth. recovery of fish not followed by the recovery of C~stoseiracanopies is a signal that MPAs are embedded within large-scale human Benthic communities impacts impairing the effectiveness of protection and requiring Benthic communities did not follow the same gradient from management interventions following an ecosystem-basedmanage- healthy to degraded associated with the enforcement of protection ment approach. in reserves. In fact, we did not find any effect of MPAs on benthic Our redundancy analysis showed that biomass of canopy algae communities, and there was no clear pattern of the structure of was negatively correlated with PP, although further analysis benthic communities associated with the gradient in fish biomass. indicated that this may be a consequence of the NW Mediterra- These results indicate that factors other than fishing are largely nean having lost most of its sublittoral Fucales historically [26]. responsible for the structure of Mediterranean benthic communi- The C~stoseiranative canopies have been disappearing throughout ties. This is not surprising since other examples from the the Mediterranean because of direct impact by fishing nets, Mediterranean Sea demonstrated that top down mechanisms are indirect impact by increase of water turbidity, habitat degradation, not always the rule within MPAs, and cascading effects are likely to urbanization and pollution [26,27,29,72]. Thus the negative vary depending on local physical conditions and on the correlation with PP we observed may result from concurrent loss characteristics of the species that are locally dominant [62]. of canopies in productive, but densely populated and developed, Unexpectedly, we did not find a strong negative correlation coastlines, not from a direct association between algal canopies between the biomass of carnivorous fishes and sea urchin biomass, and PP. as it has been found for localities situated in the northwestern The other endpoint of benthic community structure was bare Mediterranean and Adriatic, where fishes above a threshold rock, which was negatively correlated to fish biomass but weakly density appear to regulate sea urchin biomass [12]. The correlated to sea urchin biomass. This result is surprising, abundance of carnivorous fishes needed to exert top-down compared to previous research in the Western Mediterranean regulation on sea urchin populations is found only in well- where barrens are strongly correlated to sea urchin biomass, and enforced no-take reserves in the Western Mediterranean and the may be explained by the presence of barrens in the Eastern Adriatic Sea, even though this is not a general pattern [62,63]. All Mediterranean caused by alien herbivorous fishes [21]. of our Eastern Mediterranean sites had fish biomasses lower than Introduced algae were not significantly related to other algae, that threshold density of predatory fishes. Because other factors and their biomass was on average the lowest of any algal group come into play when predation is released, this could explain the throughout the Mediterranean. This is an unexpected result since wide variability in sea urchin biomass within and between sites. Mediterranean shallow water assemblagesare thought to harbor For instance, settlement of the sea urchin Paraeentrotuslividus shows the greatest number of introduced macrophytes in the world [73], strong patchiness at scales of only tens of meters [64,65]. Spatial with current estimates of 126 introduced macrophytes, 18 of them and temporal variation in recruitment may mask predatory control being considered as invasive [40]. In fact, we have detected eight of sea urchins, even in reserves with high predator biomass. There of the ten top-invasive species Aerothamnionpreissii, Asparagopsis may also be other unknown factors such as the role of PP in the sea armata, A. taxiformis, Caulerpa raeemosav. glindraeea, Halophila urchin larval pool and subsequent recruitment that could also sttpulaeea,Lophoeladia lallemandii, Sppopodiumsehimperi and Womer- determine the differences in the predatory fish sea urchin slepellasetaeea! P0,40] but almost none is dominant in any sample relationship between the eastern and the western Mediterranean. suggesting that at a Mediterranean scale and for shallow water We would expect benthic communities with large algal biomass assemblagesinvasive macrophytes may not be a major threat. to be the most mature and closer to 'pristine'. "Reference" rocky Our results also suggest that primary productivity PP! may not reefs in the Mediterranean with good water quality and without be a limiting factor in determining the maximum biomass of overgrazing by sea urchins or fish are dominated by a C~stoseira benthic communities across the Mediterranean. Availability of canopy [25,26,66]. C~stoseirabiomass was also correlated with total nutrients and irradiance regulate algal biomass throughout algal biomass except for the NW Mediterranean sites where seasonal cycles [68,74], but there were no significant differences C~stoseirahave declined historically [26]!. Therefore we expected in average algal biomass between Western and Eastern sites, C~stoseiracanopies to be indicative of 'healthy' rocky reefs. despite large differences in PP. Similarly, our overall estimates of However, most of the largest biomasses of C~stoseiracanopies cumulative human impact did not explain variation in algal where found at unprotected sites. Among the reserves with the biomass. However, results of other studies suggest that specific largest fish biomass, only Formentera-Espardell had a well impacts may affect algal biomass. The major herbivores that developed C~stoseiracanopy. The only example of recovery of a regulate algal biomass are the sea urchins Paraeentrotuslividus and

'... PLoSONE ~ www.plosone.org 10 February2012 ~ Volume 7 ~ Issue 2 ~ e32742 Mediterranean Rocky Reef Baselines

Arbaria lixula, which at large densities can create extensive barrens suitable pupping habitat and resting caves [00,03], which is denuding all non-encrusting algae [16,75]. The herbivorous fish facilitated by the presence of more than 3,000 islands and islets in Sarpasalpa can also influence the biomass and vertical distribution the Aegean Sea, most of them uninhabited. of C~stoseiraspp. [20]. In an experimental study on the coast of We did not sample the southern Mediterranean shores except Turkey, the exclusion of the alien herbivorous fishes Siganuslur dus Al-Hoceimal because of the difliculties in obtaining research and S. riuulatusresulted in an increase in algal biomass of up to 40 permits due to political issues.There are no quantitative studies on times relative to control plots within only four months [21], which the state of the nearshore rocky reef ecosystems of the southern shows that alien herbivorous fishes, and not PP or the cumulative Mediterranean. Therefore we cannot know the community human footprint, may be the major factor limiting algal growth at structure at the best preserved sites on the southern Mediterra- some Eastern Mediterranean sites. Barrens and turf algae- nean, which has the smallest number of MPAs in the dominated assemblagescaused by Siganidae [21] were common Mediterranean [84]. in our study depth range in Turkey, but below 20 25 m Siganidae were rare and erect algae were common, thus our results should be Applications to marine management restricted to the shallower depth range we investigated. Other local This study provides the first current baseline of community impacts, e.g., from coastal development and pollution, may also structure of the Mediterranean rocky sublittoral. A major insight is contribute to degrading algal assemblages[72], even within MPAs. that, at the Mediterranean scale, partially protected MPAs which allow some fishing! are not effective in restoring fish populations- Ecological baselines as opposed to well enforced no-take marine reserves, which are The difliculties of identifying appropriate reference conditions effective. Our database can be expanded, by adding comparable pose major challenges to the understanding of the causes of data from additional sites. Managers of MPAs can place their sites environmental changes [76]. There are excellent examples, on the trajectory that we have identified - from degraded Iow fish however, of reconstructed historical baselines and quantitative biomass and bare rock! to healthier large fish and algal biomass!- targets for ecosystem-basedmanagement and marine conservation and assessthe present condition of their target ecosystemsrelative P,4,77]. The use of historical baselines allows us to assessthe to this trajectory. Moreover, this trajectory yields predictions history of degradation and to guide conservation and management regarding trends to be expected during recovery, so that, in initiatives towards new ecosystem conditions having similar addition to current condition, temporal change can also be structural and functional features to those of the past. However, interpreted to determine whether recovery is occurring. For the knowledge of "pristine conditions" e.g., historical population example, monitoring a marine reserve over time and re-analyzing level or ecosystem structure, the carrying capacity under historical the data in multivariate space will allow us to determine whether ecosystem conditions! is rarely available. This reconstruction, and how protection is working at the ecosystem level and across furthermore, can be controversial in the presence of different data spatial scales. Thus, our empirically-derived gradient provides a sources or of idiosyncratic results produced by different recon- practical tool for assessing 'how your MPA is doing' from an struction methods. In the Mediterranean Sea, affected by a long ecosystem perspective. Based on the current ecosystem state, history of human-induced changes and shifting baselines, the lack management actions may be devised to promote recovery e.g., of reliable historical records represents a strong limit in setting [85]l. In addition, our results provide an additional tool and meaningful reference conditions that might assist in assessing criteria for guiding conservation planning and MPA site selection. recovery. Space-for-time substitutions may represent the only Finally, evaluation of current ecosystem state in multivariate space solution to the lack of reliable quantitative information about can inform marine management, including marine spatial historical baselines, if conclusions are derived from extensive planning, at a larger scale in situations where multiple issuesneed surveys using a consistent methodology addressing multiple sites to be addressed in order to promote recovery of rocky reef across a gradient of environmental conditions and human ecosystems. pressures, as in the present study. Additional snapshots of past conditions could surely help in refining management and Supporting Information conservation strategies. Figure Sl Relationships between pairs of species groups. It is important to note the absence of nearshore sharks at the DOCXl study sites. Sharks and other elasmobranchs were much more abundant historically in the Mediterranean P] and they used to be Table Sl List and geographic coordinates of sampling sites. an important component of nearshore food webs [2]. The largest DOCXl predators at our study sites were male dusky groupers Epinrphelus Table S2 List of taxa encountered during our quantitative raargt'natus!,which have become the dominant apex predator in surveys. Fish trophic groups: AP = Apex Predators, CA = Carni- biomass at most Mediterranean MPAs. Another apex predator, vores, Pl = Planktivores, He = Herbivores. the Mediterranean monk seal Manarhusraanarhus!, still inhabits the DOCXl Aegean Sea in Greece and Turkey, and is observed often at several of our study sites mosfly at Kimolos-Polyaigos, Karpathos, Piperi, Supporting Information Sl Cumulative human impact source Gyaros and Adrasanl [78,79]. Important reproductive groups have data and analysis. been observed at Kimolos-Polyaigos, Karpathos, Piperi and Docl Gyaros [79,00,BI]. It is remarkable that the monk seal, of which about 300 individuals remain in the Aegean, is more common at Acknowledgments the sites with lowest fish biomass. The monk seal is an The following institutions provided research permits: Medes Islands opportunistic predator that can swim over 60 km in a day and Marine Reserve,Cap de Creus Natural Park, Marine ResourcesService dive more than 100 m depth in search of prey, consuming a large and Biodiversity Department of the Balearic Islands Government, variety of food sources, although the diet of the speciesdepends to Archipelago of Cabrera National Park, Al Hoceima National Park, a large extend on cephalopods and particularly octopus [82]. The Portofino National Marine Reserve,Torre Guaceto Marine Protected major factor for M. raanarhussurvival appears to be the presence of Area, Tavolara-Punta Coda Cavallo Marine Protected Area, Porto

'... PLosONE l www.plosone.org February2012 l Volume 7 l Issue 2 l e32742 Mediterranean Rocky Reef Baselines

CesareoMarine ProtectedArea, Tremiti Marine Reserve,Capo Caccia courtesyof the Ocean Biology ProductsGroup at NASA Goddard Space Marine Protected Area, Nanonal Marine Park of Alonissos,Northern Flight Center. Pathfinder SST data were provided courtesy of NOAA Sporades,and the Ministry of ForeignAffairs of Turkey. We are indebted National OceanographicData Center and the University of Miami. to J. Amengual, the Annnea Foundation, V. Asnaghi, D. Diaz, Diving CenterAdrasan, M. Draman, S. Giakoumi,A. Grau, C.O. Kirag, M. Ko9, Author Contributions K. Kroeker, C. Linares,J.M. Llenas, A. Lorente, C. Massun, MOm- Hellenic Societyfor the Studyand Protectionof the Monk Seal,J. Moreno, Conceivedand designedthe experiments:ES EB SF PG LM FM AR MZ. P. Panzalis,D. Privitera, M. Psaradellis,V. Riera, SAD-EKOG, M. San Performedthe experiments:ES EB PD AD FF SFJG HG PG BSH BH Felix, C. Villena, and Veil Mari for logistical and field support; to Pali, AK ZK EM LM SM FM AP KR MS KS RS FT MZ. Analyzedthe dam: Nurus and Peggyfor help processingsamples; and to Ashley Greenleyfor ES EB AD FF DF SFAF PG BSH FM. Wrote the paper:ES EB FF SF PG help with analysesof remote sensingdam. MODIS data were provided BSH FM AR.

References 1. BearziG, HolcerD, SciaraGNd 004i Therole of historicaldolphin takes and of the EuropeanWater Framework Directive. Marine Pollution Bulletin 55: habitatdegradation in shapingthe present status of northernAdriatic cetaceans. 172 180. AquaticConservation: Marine and Freshwater Ecosystems 14: 363 379. 26. ThibautT, PinedoS, TorrasX, BallesterosE 005i Long-termdecline of the 2. SalaE 004i The pastand present topology and structureof mediterranean populationsof Fucales Cpstosema spp. and Sargattumspp.i in the Alberescoast subtidalrocky-shore food webs. Ecosystems 7: 333 340. France,North-western Mediterranean!. Marine Pollution Bulletin 50: 3. FerrettiF, MyersRA, SerenaF, LotzeHK008i Lossof largepredatory sharks 1472 1489. from theMediterranean Sea. Conservation Biology 22: 952 964. 27. Perkol-FinkelS,Airoldi L 010i Lossand recovery potential of marinehabitats: 4. AiroldiL, BeckM 007i Loss,status and trends for coastalmarine habitats in An experimentalstudy of factorsmaintaining resilience in subtidalalgal forests Europe.Oceanography and Marine Biology: an AnnualReview 45: 345 405. at theAdriatic Sea. PLoS ONE 5: el0791. 5. Coll M, PiroddiC, SteenbeekJ, Kaschner K, BenRais Lasram F, et al. 010i 28. SalesM, CebrianE, TomasF, BallesterosE 011i Pollutionimpacts and The biodiversityof the mediterraneansea: Estimates, patterns, and threats. recoverypotential in threespecies of the genusCpstosema Fucales, Hetero- PLoSONE 5: el 1842. kontophytai.Estuarine, Coastal and Shelf Science 92: 347 357. 6. GuidettiP, MicheliF 011i Ancientart servingmarine conservation. Frontiers 29. MangialajoL, ChiantoreM, Cattaneo-ViettiR 008i Lossof fucoidalgae along in Ecologyand the Environment 9: 374. a gradientof urbanisationand relationshipswith the structureof benthic 7. SalaE, Aburto-Oropeza0, ParedesG, RezaM, Lopez-LemusL 004i Fishing assemblages.Marine Ecology Progress Series 358: 63 74. downcoastal food webs in the Gulfof California.Fisheries 29: 19 25. 30. BoudouresqueCF, Verlaque M 002i Biologicalpollution in theMediterranean 8. JacksonJBC, SalaE 001i Unnaturaloceans. Scientia Marina 65: 273 281. Sea:invasive versus introduced macrophytes. Marine PollutionBulletin 44: 9. FriedlanderAM, DeMartiniEE 002i Contrastsin density,size, and biomass of 32 38. reeffishes between the northwestern and the main Hawaiian islands: the effects 31. RiberaMA, BoudouresqueCF 995i Introducedmarine plants, with special of fishingdown apex predators. Marine Ecology-Progress Series 230: 253 264. referenceto macroalgae:mechanisms and impact. In: RoundFE, Chapman DJ, 10. SandinS, SmithJ, DemartiniE, DinsdaleE, DonnerS, et al. 008i Baselines eds.Progress in PhycologicalResearch Biopress Ltd. pp 187 268. anddegradation of coralreefs in thenorthern Line Islands. PLoS One 3: e 1548. 32. LejeusneC, ChevaldonneP, Pergent-MartiniC, BoudouresqueCF, PerezT 11. Garcia-ChartonJA, Perez-RuzafaA, SanchezJerez P, Bayle-SempereJT, 010i Climatechange effects on a miniatureocean: the highlydiverse, highly Renones0, et al. 004i Multi-scalespatial heterogeneity, habitat structure, and impactedMediterranean Sea. Trends in Ecology& Evolution25: 250 260. the effectof marine reserveson WesternMediterranean rocky reef fish 33. Occhipinti-AmbrogiA, Galil B 010i Marinealien species as an aspect of global assemblages.Marine Biology 144: 161 182. change.Advances in Oceanographyand Limnology 1: 199 218. 12. GuidettiP, SalaE 007i Community-wideeffects of marinereserves in the 34. ClaudetJ, FraschettiS 010i Human-drivenimpacts on marinehabitats: a MediterraneanSea. Marine Ecology Progress Series 335: 43 56. regionalmeta-analysis in the MediterraneanSea. Biological Conservation 143: 13. Harmelin-VivienM, Le DireachL, Bayle-SempereJ, Charbonnel E, Garcia- 2195 2206. ChartonJA, et al. 008i Gradientsof abundanceand biomass across reserve 35. LibralatoS, Coll M, TempestaM, SantojanniA, SpotoM, et al. 010i Food- boundariesin six Mediterraneanmarine protected areas: Evidence of fish web traits of protectedand exploitedareas of the Adriatic Sea.Biological spillovergBiological Conservation 141: 1829 1839. Conservation143: 2182 2194. 14. Garcia-RubiesA, ZabalaM 990i Effectsof total fishingprohibition on the 36. MoreyG, MorantaJ, MassutiE, GrauA, LindeM, et al. 003i Weight-length rockyfish assemblagesof Medes Islands marine reserve NW Mediterranean!. relationshipsof littoral to lowerslope fishes from the westernMediterranean. ScientiaMarina 54: 317 328. FisheriesResearch 62: 89 96. 15. GuidettiP 006i Marinereserves reestablish lost predatoryinteractions and 37. FroeseR, PaulyD 011i FishBase.World Wide Web electronic publication. causecommunity changes in rockyreefs. Ecological Applications 16: 963 976. www.fishbase.org. 16. SalaE, BoudouresqueCF, Harmelin-Vivien M 998i Fishing,trophic cascades, 38. NewmanMJH, Paredes GA, Sala E,Jackson JBC 006i Structureof Caribbean and the structureof algal assemblages:evaluation of an old but untested coralreef communities across a largegradient of fishbiomass. Ecology Letters 9: paradigm.Oikos 82: 425139. 1216 1227. 17. CardonaL, SalesM, LopezD 007i Changesin fishabundance do not cascade 39. StergiouKI, Karpouzi VS 001i Feedinghabits and trophic levelsof to seaurchins and erectalgae in one of the mostoligotrophic parts of the Mediterraneanfish. Reviews in FishBiology and Fisheries 11: 217 254. Mediterranean.Estuarine, Coastal and Shelf Science 72: 273 282. 40. ZenetosA, GofasS, VerlaqueM, QinarE, Garcia-RasoE, et al. 010i Alien 18. AzzurroE, FanelliE, MostardaE, Catra M, AndaloroF 007i Resource speciesin the Mediterraneanareas of the EuropeanUnion's Marine Strategy partitioningamong early colonizing S>gauus luudus and native herbivorous fish in FrameworkDirective MFSDi by 2010.Part I. Spatialdistribution. Mediterra- theMediterranean: an integrated study based on gut-contentanalysis and stable neanMarine Science 11: 38IW93. isotopesignatures. Journal Marine Biological Association United Kingdom 87: 41. BallesterosE 992i Elsvegetals i la zonaciolitorak cspecies, comunitats i factors 991 998. queinflueixen en la sevadistribucio. Barcelona: Institut d'Estudis Catalans. 19. TomasF, CebrianE, BallesterosE 011i Differentialherbivory of invasivealgae 42. PaisA, ChessaLA, SerraS, Ruiu A, MeloniG, et al. 007i The impactof by nativefish: temporaland spatialvariability. Estuarine, Coastal and Shelf commercialand recreationalharvesting for Paracentrotuslividus on shallow Science92: 27 34. rockyreef sea urchin communities in North-westernSardinia, Italy. Estuarine, 20. VergesA, AlcoverroT, BallesterosE 009i Roleof fishherbivory in structuring Coastaland Shelf Science 73: 589 597. thevertical distribution of canopyalgae Cystoseira spp.i in the Mediterranean 43. FriedlanderA, BrownE, MonacoM 007i Definingreef fish habitat utilization Sea.Marine Ecology Progress Series 375: 1 11. patternsin Hawaii:comparisons between marine protected areas and areas open 21. SalaE, KizilkayaZ, YildirimD, BallesterosE 011i Alienmarine fishes deplete to fishing.Marine Ecology Progress Series 351: 221 233. algalbiomass in theEastern Mediterranean. PLoS ONE 6: el 7356. 44. RiskM 972i Fishdiversity on a coralreef in theVirgin Islands. Atoll Research 22. ZabalaM, BallesterosE 989i Surface-dependentstrategies and energy flux in Bulletin153: I&. benthicmarine communities or, whycorals do not existin theMediterranean. 45. CaseyK, BrandonT, CornillonP, Evans R 010i ThePast, Present and Future ScientiaMarina 53: 1 15. of theAVHRR Pathfinder SST Program. In: BaraleV, GowerJ, AlberotanzaL, 23. SalesM, BallesterosE 009i Shallow Cpstosema Fucales: Ochrophytai eds.Oceanography from Space:Revisited: Springer, DOI: 10.1007/978-90- assemblagesthriving in shelteredareas from Menorca NW Mediterranean!: 481-8681-516. relationshipswith environmentalfactors and anthropogenicpressures. Estua- 46. BehrenfeldM, FalkowskiP 997i Photosyntheticrates derived from satellite- rine,Coastal and Shelf Science 84: 476 482. basedchlorophyll concentration. Limnology and Oceanography 42: 1 20. 24. SalesM, BallesterosE 010i Long-termcomparison of assemblagesdominated 47. HalpernBS, Walbridge S, SelkoeKA, KappelCV, MicheliF, et al. 008i A by Cpstosemacuruta Bory Fucales,Ochrophytal from Cap Corse Corsica, GlobalMap of HumanImpact on MarineEcosystems. Science 319: 948 952. NorthwesternMediterranean!. European Journal of Phycology45: 404 412. 48. GuidettiP, MilazzoM, BussottiS, Molinari A, MurenuM, et al. 008i Italian 25. BallesterosE, TorrasX, PinedoS, Garcia M, MangialajoL, et al. 007i A new marinereserve effectiveness: Does enforcement matterg Biological Conservation methodologybased on littoralcommunity cartography for the implementation 141:699 709.

'... PLoSONE ~ www.plosone.org 12 February2012 ~ Volume 7 ~ Issue 2 ~ e32742 Mediterranean Rocky Reef Baselines

49. AndersonM 001l A newmethod for non-parametricmultivariate analysis of 68. HereuB, ZabalaM, SalaE 008l Multiplecontrols of communitystructure and variance.Austral Ecology 26: 3216. dynamicsin a sublittoralmarine environment. Ecology 89: 3423 3435. 50. PaulyD, ChristensenV, DalsgaardJ, FroeseR, TorresF 998l Fishingdown 69. MangialajoL, ChiantoreM, SusiniM-L, MeineszA, Cattaneo-ViettiR, et al. marinefood webs. Science 279: 860 863. 011l Zonationpatterns and interspecific relationships of fucoids in microtidal 51. Micheli F, HalpernBS, BotsfordLW, WarnerRR 004l Trajectoriesand environments.Journal of ExperimentalMarine Biology and Ecology in press!. correlatesof communitychange in no-takemarine reserves.Ecological 70. FrancourP 994lPluriannual analysis of thereserve effect on the ichtyofauna in Applications14: 1709 1723. the Scandolanatural reserve Corsica, northwestern Mediterranean!. Oceano- 52. JongmanR, ter BraakC, Van Tongeren0, eds.Data analysis in community logicaActa 1 7:309 31 7. andlandscape ecology. Cambridge: Cambridge University Press. 299 p. 71. BallesterosE, HereuB, ZabalaM, AlcoverroT, GarrabouJ, et al. 002l 53. BraakCt, SmilauerP 002l CANOCOReference Manual and CanoDraw for Rapportmission scandola: Cystoseira 2000. Travaux Scientifiques du Pare WindowsUser's Guide: Software for CanonicalCommunity Ordination version NaturelRegional de la Corse60: 95 115. 4.5l. Sectionon permutationmethods. Microcomputer Power IthacaNY, 72. MangialajoL, RuggieriN, AsnaghiV, ChiantoreM, PoveroP, et al. 007l USAi.500 p. Ecologicalstatus in theLigurian Sea: the effect of coastlineurbanisation and the 54. BraakCt 994l Canonicalcommunity ordination. Part I: Basictheory and importanceof properreference sites. Mar PollutBull 55: 30 41. linearmethods. Ecoscience 1: 127 140. 73. KleinJ, RuittonS, Verlaque M, BoudouresqueC 005l Speciesintroductions, 55. BraakCt, VerdonschotP 995l Canonicalcorrespondence analysis and related diversityand disturbances in marine macrophyte assemblages of the northwest- multivariatemethods in aquaticecology. Aquatic Sciences 55: 1 35. ernMediterranean Sea. Marine Ecology Progress Series 290: 79 88. 56. SalaE 997l Fishpredators and scavengersof the seaurchin Paracentrotus 74. BallesterosE 989l Productionof seaweedsin NorthwesternMediterranean lividusin protectedareas of thenorth-west Mediterranean Sea. Marine Biology marinecommunities: Its relationwith environmentalfactors. Scientia Marina 129:531 539. 53: 357 364. 57. DeMartiniE, FriedlanderA, SandinS, SalaE 008l Differencesin fish 75. GuidettiP, FraschettiS, Terlizzi A, BoeroF 003l Distributionpatterns of sea assemblagestructure between fished and unfishedatolls in the northernLine urchinsand barrens in shallowMediterranean rocky reefs impacted by theillegal Islands,central Pacific. Marine Ecology Progress Series 365: 199 215. fisheryof the rock-boringmollusc Lirhsphaga Iirhophaga. Marine Biology 143: 58. BallesterosE, Garcia-RubiesA, Mariani S, ComaR, Diaz D, et al. 008l 1135 1142. Monitoringof theMarine Protected Areas of Capde Creus and Medes Islands. 76. Bulleri F, UnderwoodA, Benedetti-CecchiL 007l The assessmentand TechnicalReport to theCatalan Government, Departament de Medi Ambient i interpretationof ecologicalimpacts in human-dominatedenvironments. Habitatge.80 p. EnvironmentalConservation 34: 181 182. 59. Aburto-Oropeza0, ErismanB, GallandGR, Mascarenas-OsorioI, Sala E, 77. LotzeHK, LenihanHS, BourqueBJ, Bradbury RH, CookeRG, et al. 006l et al. 011l Largerecovery of fishbiomass in a no-takemarine reserve. PLoS Depletion,Degradation, and Recovery Potential of Estuariesand Coastal Seas. ONE 6: e23601. Science312: 1806 1809. 60. McClanahanTR, GrahamNAJ, CalnanJM, MacNeilMA 007l Toward 78. GuclusoyH, Kirac,C, VeryeriN, SavasY 004l Statusof the Mediterranean pristinebiomass: Reef fish recoveryin coral reefmarine protected areas in monkseal, Monarhus moriarhus Hermann, 1779l in thecoastal waters of Turkey. Kenya.Ecological Applications 17: 1055 1067. Journalof Fisheries& AquaticSciences 21: 201 210. 61. Di FrancoA, BussottiS, NavoneA, PanzalisP, GuidettiP 009l Evaluating 79. MOm 009l Statusof the Mediterraneanmonk sealMonarhus moriarhus in effectsof totaland partial restrictions to fishingon Mediterranean rocky-reef fish Greece.Athens: MOm HellenicSociety for the Studyand Protection of the assemblages.Marine Ecology Progress Series 387: 275 285. MonkSeal. 36 p. 62. MicheliF, Benedetti-CecchiL, Gambaccini S, Bertocci I, BorsiniC, et al.005l 80. DendrinosP, KaramanlidisA, KotomatasS, Legakis A, TountaE, et al. 007l Cascadinghuman impacts, marine protected areas, and the structureof Puppinghabitat use in theMediterranean monk seal: a long-termstudy. Marine mediterraneanreef assemblages. Ecological Monographs 75: 81 102. MammalScience 23: 615&28. 63. GuidettiP, VerginellaL, Viva C, OdoricoR, BoeroF 005l Protectioneffects 81. DendrinosP, KaramanlidisA, KotomatasS, ParavasV, AdamantopoulouS on fishassemblages, and comparison of twovisual-census techniques in shallow 008l Reportof a NewMediterranean monk seal Monarhusmsnarhus! breeding artificialrocky habitats in the northernAdriatic Sea. Journal of the Marine colonyin theAegean Sea, Greece. Aquatic Mammals 34: 355 361. BiologicalAssociation of the UnitedKingdom 85: 247 255. 82. PierceG, Hernandez-MilianG, SantosM, PsaradellisM, TountaE, et al.011l 64. HereuB, LinaresC, ZabalaM, SalaE 004l Temporaland spatial variability in Dietof themonk seal Monarhus monarhusl in Greek waters. Aquatic Mammals 27: setdementof thesea urchin Parazentrstus lividus Lamarckl. Marine Biology 144: 284 297. 1011 1018. 83. JohnsonW, KaramanlidisA, DendrinosP, Larrinoa PFd, Gazo M, et al. 006l 65. TomasF, RomeroJ, TuronX 004l Settlementand recruitmentof the sea Monk sealfact files. biology,behaviour, status and conservationof the urchin Parazentrotuslividus in two contrastinghabitats in the Mediterranean. Mediterraneanmonk seal,Monarhus msriarhus. The MonachusGuardian, MarineEcology Progress Series 282: 173 184. wwwmonachus-guardianorg. 66. PinedoS, GarciaM, SattaM, de TorresM, BallesterosE 007l Rocky-shore 84. AbdullaA, GomeiM, HyrenbachD, Notarbartolo-di-SciaraG, AgardyT communitiesas indicators of waterquality: a casestudy in the Northwestern 009l Challengesfacing a networkof representativemarine protected areas in Mediterranean.Marine Pollution Bulletin 55: 126 135. the Mediterranean:prioritizing the protectionof underrepresentedhabitats. 67. SalaE, BoudouresqueCF 997l The role of fishesin the organizationof a ICESJournal of MarineScience 66: 22 28. Mediterraneansublittoral community.: I: Algal communities.Journal of 85. GuidettiP, ClaudetJ 010l Co-managementpractices enhance fisheries in ExperimentalMarine Biology and Ecology 212: 25 44. marineprotected areas. Conservation Biology 24: 312 318.

'... PLoSONE ~ www.plosone.org 13 February2012 ~ Volume 7 ~ Issue 2 ~ e32742