Climate‐driven changes in macrobenthic communities in the Mediterranean Sea: A 10‐year study in the Bay of Banyuls‐sur‐Mer

Abstract Marine ecosystems worldwide are affected by both natural variation and human activities; to disentangle and understand their individual role in influencing the macrobenthic community composition is challenging. The relationship between interannual variability in atmospheric circulation, dictated by the climatic oscillation indices, and the benthic macrofauna composition was assessed at four sampling sites located in the Bay of Banyuls‐sur‐Mer (NW Mediterranean Sea). Between 2004 and 2013, these sites were sampled annually during autumn/winter and analyzed for sediment grain‐size and benthic macrofauna composition (species richness, abundance, and biomass). Temporal changes in these descriptors were correlated with two climatic indices (NAO and WeMO indices) and a set of environmental parameters integrated over three different time periods (i.e., whole year, springtime, and wintertime). Our results confirm the occurrence of major temporal changes in the composition of macrobenthic communities within the Gulf of Lions. More specifically, the results indicate that (a) the WeMO appears to be more closely related to benthic macrofauna composition in the Bay of Banyuls‐sur‐Mer than the NAO, (b) winter is a better integration period than spring or the whole year as a proxy for community composition changes, and (c) Rhône River water flow is likely involved in the control of benthic macrofauna composition in the whole Gulf of Lions. The present study highlights the importance of WeMO as a regional proxy, which can be used to evaluate changes in benthic macrofauna linked to climatic variability.

is the major source of interannual variability in the atmospheric circulation in the North Atlantic (Hurell, 1995). The NAO largely controls local changes in a large set of meteorological parameters such as water temperature, salinity, wind strength/direction, and storms.
The NAO index provides a good summary of general weather patterns influencing marine ecosystems and affecting the abundance, biomass, growth, and survival rates of marine organisms (Drinkwater et al., 2003;Fromentin & Planque, 1996;Kröncke, Dippner, Heyen, & Zeiss, 1998;Shojaei et al., 2016). Several studies have highlighted the consequences of changes in meteorological parameters and, thus, of NAO on (a) zooplankton communities in the western Mediterranean (Fernández de Puelles, Valencia, & Vincent, 2004); the North Atlantic and the North Sea (Fromentin & Planque, 1996); (b) fisheries in the NW Mediterranean Sea (Lloret, Lleonart, Sole, & Fromentin, 2001); (c) recruitment of anchovy (Santojanni et al., 2006) in the Adriatic Sea; (d) physical condition of migratory bullet tuna stock during pre-and postreproductive movement (Muñoz-Expósito et al., 2017) in the western Mediterranean; and (e) benthic macrofauna composition in the North Sea (Hagberg & Tunberg, 2000;Kröncke et al., 1998Kröncke et al., , 2011Kröncke, Zeiss, & Rensing, 2001;Rees et al., 2006;Shojaei et al., 2016;Tunberg & Nelson, 1998). Between 1978and 1995, Kröncke et al. (1998 seasonally sampled five sites located between 12 and 20 m depth off the Island of Norderney. The study demonstrated that the abundance and species richness of benthic macrofauna sampled between April and July were correlated significantly with the NAO index. The authors suggested that the mediator between NAO and benthic macrofauna was sea surface temperature (SST) in late winter and early spring.
This SST-driven hypothesis was supported by further observations by Beukema (1985) who reported the decrease in Echinocardium cordatum populations after severe winters. Lloret et al. (2001) were the first to correlate climatic oscillations with biological parameters in the NW Mediterranean and studied the relationship between fish and invertebrate landings, the Rhône and Ebre River water flow, and the NAO index.
Findings from the study included (a) a negative correlation between water flow of these two rivers and the NAO, and (b) a positive correlation between the landings of 13 species of fishes and invertebrates and water flow. The authors suggested a link between recruitment and local environmental conditions such as river discharge, wind speed and direction, and global environmental conditions (i.e., NAO). However, recent studies in the NW Mediterranean have focused on the Western Mediterranean Oscillation index (WeMO index) rather than the commonly used NAO index as a proxy of local climatic variability (Martín, Sabatés, Lloret, & Martin-Vide, 2012;Martin-Vide & Lopez-Bustins, 2006;Martin-Vide et al., 2008). These two indices do not correlate significantly when standardized on an annual basis or in wintertime (Martín et al., 2012;Martin-Vide & Lopez-Bustins, 2006). The WeMO index has been shown to be more relevant than the NAO index to account for monthly precipitation anomalies in the Iberian Peninsula (Martin-Vide & Lopez-Bustins, 2006;Martin-Vide et al., 2008). Further, Martín et al. (2012) showed that positive WeMO index values correlated significantly with low SST and high river run-offs, which have a significant positive effect on sardine and anchovy landings per unit effort. Conversely, and based on a 45year time series, Keller, Valls, Hidalgo, and Quetglas, (2014) did not show any influence on the landings of Sepia officinalis in the western Mediterranean by either the NAO or the WeMO index, but only by SST.
Most of the benthic macrofauna data available in the Gulf of Lions (Bonifácio et al., 2018;Grémare, Amouroux, & Vétion, 1998;Grémare, Sardá, et al., 1998;Labrune, Grémare, Guizien, & Amouroux, 2007;Labrune et al., 2008;Massé, 2000;Salen-Picard, 1981) have been collected over too narrow time scales to soundly assess their correlation with climatic oscillations whereas few studies by achieving long-term comparisons. In 1967/68, Guille (1970) first described the benthic macrofauna communities of soft-bottom habitats of the Catalan French coast. Grémare, Amouroux, and Vétion (1998) then demonstrated the occurrence of major changes in both sediment grain-size and macrofauna composition between 1967/68 and 1994, and suggested that these changes were due to the decrease in fine particles most likely caused by an increase in the frequency of easterly storms. By using a different procedure to assess resuspension events, Labrune, Grémare, Guizien, et al. (2007) suggested that positive NAO index periods were related to low frequency of strong resuspension events and high abundance and biomass of benthic fauna. The underlying hypothesis put forward from this work was that the low frequency of resuspension events, especially during springtime, con- inferred that NAO may have more influence on the recruitment of D. arietina than WeMO in the Gulf; but, they highlighted the urgent need for a long-term monitoring series in order to confirm their hypothesis. In this context, the main objective of this study was (a) to assess changes in sediment grain-size and benthic macrofauna composition based on data collected annually from 2004 to 2013 in the Bay of Banyuls-sur-Mer and (b) to evaluate the relationship between these changes with NAO and WeMO indices, and the main environmental parameters which are affecting the NW Mediterranean Sea. To achieve these objectives, the present study focuses on benthic macrofauna composition at four sampling sites, which are representative of the main benthic communities described by Guille (1970).

| Study area and sampling sites
The Bay of Banyuls-sur-Mer is located within the Gulf of Lions in the northwestern Mediterranean Sea (Figure 1). Within this bay, four sites were sampled once a year between 2004 and 2013 (Table 1).
Sites were chosen to represent the main benthic communities described by Guille in 1968(Guille, 1970Labrune, Grémare, Guizien, et al., 2007), whereby community names were maintained in accordance with those described by Guille (1970). Sites were sampled for sediment grain-size analysis and benthic macrofauna during the end of autumn/beginning of winter (November-December) on board the RV Nereis II.

| Grain-size analysis
At each sampling site, a 0.1 m 2 van Veen grab was taken for sediment grain-size analysis. Sediment grain-size analysis was performed on fresh sediment using a Malvern Mastersizer ® 2000 laser microgranulometer and expressed as median grain diameter (D 0.5 ) and in volume percentages of grain-size fractions (<30, 30-63, 63-250, 250-500, 500-2,000 µm

| Benthic macrofauna
Five replicate grab samples were taken per sites for faunal analysis, immediately sieved on a 1 mm mesh, and fixed with 5% formalin buffered in seawater. At the laboratory, macrofauna was sorted, identified to the lowest possible taxonomic level (most often to species level), and counted. Biomass was assessed by measuring the weight loss after combustion (450°C, 5 hr) of dried samples.

| North Atlantic Oscillation
The North Atlantic Oscillation is responsible for changes in the trajectories of surface westerlies across the North Atlantic toward Europe (Hurrell, 1995). Such changes can be described through several indices of NAO estimated using different approaches. During the present study, we used the classical NAO index developed by Hurrell and Deser (2009) based on the principal component (PC) time series of the leading empirical orthogonal function (EOF) of Sea Level Pressure anomalies over the Atlantic area (20°-80°N, 90°W-40°E). This method presents better representations of the full spatial patterns of the NAO. Positive values are typically associated with stronger-than-average westerlies and storms over northern Europe and milder weather with less-than-average storms over western Europe and the Mediterranean Sea. Corresponding data were provided by the Climate Analysis Section (NCAR, Boulder, USA, https ://clima tedat aguide.ucar.edu/clima te-data/hurre ll-northatlan tic-oscil lation-nao-index-pc-based ).

| Western Mediterranean Oscillation
The WeMO is a low-frequency variability pattern of atmospheric circulation that was first described by Martin-Vide and Lopez-Bustins Banyuls-sur-Mer, the negative phase is therefore associated with easterlies, which lead to frequent resuspension events. We used WeMO index data from http://www.ub.edu/gc/Engli sh/wemo.htm.
For the NAO and WeMO indices, the winter value for year n corresponds to an average from December year n-1 to February year n.
Annual and spring values corresponded to the average of monthly values from January to December and from March to May, respectively.

| Environmental parameters
Water flow of the Rhône River was provided by Banque Hydro (http:// www.hydro.eaufr ance.fr). Air temperature, precipitation, wind speed, (http://somlit.epoc.u-borde aux1.fr/fr). Criteria 2 (C2) proposed by Labrune, Grémare, Guizien, et al. (2007) was used as a proxy for intense resuspension events. In brief, C2 corresponds to an estimated number of resuspension events per year. An intense resuspension event was assumed to take place during each day featuring both a wind direction between 90° and 170° and a decrease in SLP higher than 5 hPa either between (a) the day before and the day of measurement or (b) the day of measurement and the day after. For all parameters, seasonal values were computed as described above for NAO and WeMO indices.

| Benthic macrofauna
To enable better community descriptions and allow comparison with previous studies in the area, data from replicate grabs per sampling site were pooled (Ellingsen, 2001). The results of data analyzed by averaging replicates per site are available in Supplementary material (i.e., Tables S1-S4 corresponding to Tables 3-6; and Figures S1-S3 corresponding to Figures 5-7). When possible, taxa were identified to species level, but taken to a higher taxonomic level when confidence was low, thereby allowing species data to be comparable across datasets (i.e., from different years). Synonyms of scientific names of species were updated using the World Register of Marine Species whether multivariate within-group dispersion was homogenous for sampling site groups using the PERMDISP procedure (Anderson, 2001(Anderson, , 2006. SIMilarity PERcentage (SIMPER) analysis (Clarke, Somerfield, & Gorley, 2008) was performed to identify the species contributing most to dissimilarity between subclusters.

| Relationships linking climatic variability, environmental parameters, and benthic macrofauna
All analyses relating climatic variability and environmental parameters were run using benthic macrofauna by site rather than TA B L E 3 Global descriptors of sediment grain-size (D 0.5 in µm and proportion of fine sediment in % <63 µm) and benthic macrofauna composition (species richness in taxa.0.5 m −2 , abundance in ind.0.5 m −2 , and biomass in mgAFDW.0.5 m −2 )  all analyses were completed using the PRIMER 6 ® software package (Clarke & Gorley, 2006;Clarke & Warwick, 2001).
TA B L E 4 Abundance, contribution, and cumulative contributions to dissimilarities in benthic macrofauna composition for the five species most responsible for dissimilarity between the subclusters identified in Figure 6b Subclusters

| Temporal changes in climatic indices and environmental parameters
Strong temporal changes in annual, spring, and winter values of the

| Temporal changes in benthic macrofauna
In total, 15,431 specimens belonging to 448 taxa were identified. An overarching pattern was observed for temporal changes in macrobenthos species richness and abundance (Table 3 and    The species most responsible for dissimilarity between subcluster located at shallower waters were Ditrupa arietina and Aspidosiphon muelleri (clusters I and II), whereas those responsible for dissimilarity between subcluster deeper water (cluster III) were A. muelleri and Turritella communis (Table 4).

| Relationship between climatic indices, environmental parameters, sediment grain-size, and benthic macrofauna
No significant correlation between the annual or winter NAO index values and the global descriptor of benthic macrofauna was found for any sampling site (Table 5) (Table 5)   All these studies, however, were based on either long-term comparison data collected during specific time periods with long intervals (Bonifácio et al., 2018;Grémare, Amouroux, & Vétion, 1998;Labrune, Grémare, Guizien, et al., 2007) or on indicator species (Grémare, Sardá, et al., 1998;Medernach et al., 2000), which clearly complicates ecological interpretations and reduces the strength of derived conclusions (Grémare, Amouroux, & Vétion, 1998;Pearson, Josefson, & Rosenberg, 1985;Rosenberg, Gray, Josefson, & Pearson, 1987) when compared to studies based on long time series and/or the analysis of whole community composition (Hewitt et al., 2016;Kröncke et al., 1998Kröncke et al., , 2011Kröncke et al., , 2001. Within this context, the present study consisted of the acquisition and analysis of long time series collected within the Bay of Banyulssur-Mer at four sampling sites (located between 15 and 43 meters depth), which are representative of the main benthic communities described in this area (Guille, 1970).

| Temporal changes in benthic macrofauna composition and potential indicator species
Our results first confirm the occurrence of major temporal (i.e., interannual) changes in the composition of benthic macrofauna at the four studied sites and showed that recent changes were most  Labrune, Grémare, Guizien, et al. (2007). Conversely, they do not support the fact that these changes tend to be more important at site 31 (26 m depth) than at sites 26 (31 m depth) and 183 (43 m depth) in Labrune, Grémare, Guizien, et al. (2007). by previous studies (Bonifácio et al., 2018;Grémare, Sardá, et al., 1998;Labrune, Grémare, Guizien, et al., 2007), which is confirmed by the present study. Another potential indicator species at site 31, already identified by Labrune, Grémare, Guizien, et al. (2007), is the sipunculid Aspidosiphon muelleri. Our data also lead to the identifi-

| NAO and WeMO indices, integration periods
Two of the major results from the present study were the lack of  (Bonifácio et al., 2018;Labrune, Grémare, Guizien, et al., 2007).
Regarding the first result, NAO and WeMO indices can both be seen as proxies for changes in environmental parameters (Drinkwater et al., 2003;Lockerbie, Coll, Shannon, & Jarre, 2017;Ottersen et al., 2001). Mer, see Figure 1), the water flow of Rhône was measured close the mouth and it accounts for the entire hydrological basin (97,800 km 2 ) which is submitted to strong climatic heterogeneity (Pont, Simonnet, & Walter, 2002).
Regarding the most appropriate integration period, it was first suggested that NAO has a stronger control on the climate of the Northern Hemisphere during wintertime. During this season, the magnitude and spatial coherence of atmospheric circulation variability as well as the influence of circulation changes and large-scale precipitation are stronger (Osborn, 2006;Osborn, Conway, Hulme, Gregory, & Jones, 1999 (Kröncke et al., 1998(Kröncke et al., , 2001Kröncke, Reiss, & Dippner, 2013), in the German Bight (Neumann, Ehrich, & Kröncke, 2008;Shojaei et al., 2016), and in the Wadden Sea (Beukema, Essink, & Dekker, 2000

| Key environmental parameters involved
In the North Sea, the key environmental parameter controlling benthic macrofauna composition is wintertime temperature (Beukema et al., 2000;Kröncke et al., 1998Kröncke et al., , 2013Kröncke et al., , 2001Neumann et al., 2008;Shojaei et al., 2016). Our results suggest similar processes occurring in the NW Mediterranean under hydro-sedimentary forcing.
First, two of the potential indicator species identified during the present study (i.e., Ditrupa arietina and Turritella communis) are both suspension-feeders. Turritella communis can remain buried in mud filtering for long period unless disturbed (Yonge, 1946). Yonge reported that T. communis is very sensitive to SPM and that it stops its inhalant current as soon as fine sediment enters in its mantle cavity. Consequently, the negative phase of wintertime WeMO index, related with a high frequency of resuspension events, may have a strong impact on the population dynamic of this particular species.
Our results suggest that the mediators between winter WeMO and benthic macrofauna (based on abundances at sampling sites 31, 26, and 183 and biomasses at sampling site 31) were as follows: winter precipitation, winter Rhône River water flow, winter C2, winter SPM, and winter wind speed. This is in accordance with Lloret et al. (2001) who suggest a link between recruitment and local environmental conditions such as river discharge, wind speed, and direction and climatic oscillation. Interestingly, most of these parameters are linked with hydro-sedimentary processes. Suspended Particulate Matter, C2, and wind speed are all directly or indirectly related to sediment resuspension within the Bay of Banyuls-sur-Mer (Ferré et al., 2005;Grémare et al., 2003;Labrune, Grémare, Guizien, et al., 2007), whereas the Rhône River is the main source of continental particles for the Gulf of Lions (Durrieu de Madron et al., 2000). The impact of the Rhône River on benthic macrofauna composition has already been described in the immediate vicinity of the Rhône River mouth (Bonifácio et al., 2014(Bonifácio et al., , 2018Darnaude, Salen-Picard, Polunin, & Harmelin-Vivien, 2004;Salen-Picard, Arlhac, & Alliot, 2003;Salen-Picard et al., 2003).   Nevertheless, and because it is the first study to involve the acquisition of long-term time series of benthic macrofauna in the Catalan Sea, our study allows the paradigm to be refined by concluding that (a) the WeMO appears to be more closely related than NAO to benthic macrofauna composition in the Bay of Banyuls-sur-Mer, (b) winter is a better integration period than spring or the whole year to be used as a proxy in community composition changes, and finally, (c) Rhône River water flow is probably involved in the control of benthic macrofauna composition in the whole Gulf of Lions.

ACK N OWLED G M ENTS
We thank the captain and crew of the RV Nereis II for technical as- Grall, Rafael Sardá, Johan Eklöf (Associate Editor), and the anonymous reviewers for their helpful comments.

CO N FLI C T O F I NTE R E S T
None declared.