The study was conducted on a muddy fishing ground located in the northwestern Mediterranean. This fishing ground spreads over 400 km² in a depth range between 30 and 80 m, and the study area covers a depth range of 40-60 m (Fig. 2). Within this study area, samples were taken from a fished site and from a control site that had remained undisturbed for 20 years due to the presence of the remains of an oil platform (Fig. 2) (see de Juan et al. 2007b for details). The fishing ground was operated by Sant Carles de la Ràpita trawling fleet that, with 69 vessels, is the most important trawling fleet in the Catalonia region (northeast Spain). This fleet showed variable activity throughout the year: fishing effort was high during autumn and winter, low in spring and at the beginning of summer and observed a closed season during July and August. Data on fishing effort was obtained from St. Carles de la Ràpita fishermen’s association (see Demestre et al., 2008 for further details). Several abiotic characteristics of the area were measured during the benthic sampling, which was timed to cover seasonal pattern of fishing effort (Table 1). The mud sediment content was almost 100% over the whole study period and the temporal variability was not significant, which characterizes the area as a muddy habitat.

Samples of epifauna and infauna were collected during seven experimental cruises. Epifauna was collected with a surface dredge, similar to a 2 m beam-trawl with a 1-cm cod-end, and infauna with a 0.1-m² Van Veen grab. On each cruise, a total of three epifaunal and five infaunal replicates were randomly collected at both fished and control sites. To collect the minimum sample size, estimated from species accumulation curves, the surface dredge was towed for approximately 15 minutes for each replicate and five grabs were collected per replicate. Epifaunal and infaunal organisms were identified to the lowest possible taxonomical level, generally species for epifauna and genera for infauna, and counted (see de Juan et al. 2007b for details).

Data on landings and income from 2000 to 2011 were obtained from records from the local fish auction that takes place upon the arrival of vessels at port (data source: fishing statistics elaborated by the Fisheries Department of the Catalan government). Data were available on daily landings by species (weight and income) for each fishing vessel.

Trait classification to characterize benthic communities

Eleven biological traits covering aspects of the benthic organisms’ morphology, feeding patterns and life histories were selected to represent benthic community. The biological trait approach (BTA) allows community structure and functionality to be better represented in order to link them to the ecosystem services that the community can provide. In our case, a benthic ecosystem from a fishing ground, one of the main ecosystem services that this community provides is food production.

These 11 traits were broken down into categories. For example, feeding type was separated into the categories deposit feeder, filter/suspension feeder, opportunist/scavenger and predator (Table 2). The trait "age at sexual maturity" was treated differently for infauna and epifauna data due to the different life histories of these two groups, with most of the infauna taxa reaching the sexual maturity before a year time and epifauna species maturing later. To reduce the task to a manageable size, the data sets were reduced. For the infaunal assemblage, 25 of 147 taxa were used, comprising the species that contributed 80% of the total abundance plus those that, though not among the most abundant species, were continuously present in three of the five replicates either at fished or control sites over the study period. The epifaunal data set was reduced to 17 of 96 species, which accounted for 95% of the total abundance and met the frequency of occurrence criteria.

Each taxon in the database was scored for its affinity to each trait category using a scale of 0–3 (0 = no affinity to 3 = high affinity). The score was given using the ‘fuzzy scoring’ method, which allowed the taxa to exhibit more than one category of a given trait as long as the total score per trait was 3 (Bremner et al., 2003). This assignment was based on published accounts of the biology of each species and information codified in the BIOTIC database maintained by the Marine Biological Association UK (http://www.marlin.ac.uk/biotic/). This source of information was complemented with a literature review for potential regional differences, i.e. studies conducted in the Mediterranean. When information was not available at the species level we used data based on accounts of other members of the genera or, rarely, family (only in 7.6% of cases for infauna and 3.2% for epifauna). When no information on a particular trait was available for a taxon, zero values were entered for each category and the taxon did not contribute to the calculation of trait weightings.

The frequency of each trait category in the dataset was calculated by weighting the category scores by the abundance (number of individuals per m²) of each taxon exhibiting that category (Charvet et al. 1998). This resulted in a sample by trait table that showed the abundance of biological traits at each station over the study period.

Similarity between each pair of samples was calculated using the Bray-Curtis index after a square root transformation of the data to reduce the influence of dominant traits/species. A PERMANOVA analysis was used to test for significant differences between sites (control and fished as fixed factors) and time (fishing effort periods, i.e. before, during and after closure, as fixed factor). The trait data were further analysed with the SIMPER routine to determine which traits accounted for the significant dissimilarities identified by PERMANOVA. Then, the most important traits highlighted by SIMPER (traits showing ratio of dissimilarity to standard deviation [diss/sd] >1.5 and being among the ones summing 50% of cumulative contribution to dissimilarity) were selected for univariate analyses. When traits had a normal distribution and homogeneity of variances, a two-way ANOVA was performed to test for the factors treatment and effort period. If traits were not normally distributed (even after log transformation), a Kruskall-Wallis test was performed instead. All the multivariate analyses were carried out using the PRIMER6 & PERMANOVA statistical package (Anderson et al. 2008). Univariate analyses were performed using the R program, v.2.11.0.

Figure 3 shows the percentage of mantis shrimp (Squilla mantis), hake (Merluccius merluccius) and red mullet (Mullus barbatus) landings and income with respect to the total landings and income in the study area. The importance of these three species remained consistent over the years, representing around 25% of the total landings and income. The "other" percentage comprises on average 53 species, most of them accounting for less than 5% of the total catches and only sporadically (e.g. in one or two years of the data series) more than 5%.

Despite an overall decrease since 2007, red mullet landings consistently followed the same trend, with a peak of catches in September/October (the recruitment months for this species), which accounted for almost 20% of the total landings (Fig. 4A, B). Figure 4C shows how this peak in red mullet catches coincided with the high effort period. However, red mullet landings dropped sharply in late autumn (November) whereas the decreasing trend of fishing effort was smoother as mantis shrimp landings were maintained over the winter (Fig. 4C). This figure shows how fleet dynamics followed red mullet population, as the highest landings and effort occurred just after the closed season, when the trawling fleet gathered on this fishing ground to fish red mullet recruits (Demestre et al. 1997, Martín et al. 1999).

PERMANOVA analyses highlighted significant differences for both infauna and epifauna between sites (control vs. fished) and time (different effort regimes), and site:time interaction for infauna (Table 3). Pairwise tests performed for infauna within fished sites showed significant differences between before closure and the closed season (p<0.01) and between before closure and after closure periods (p<0.05). Furthermore, differences between times for epifauna were independent of the site, and pairwise tests performed across all samples showed significant differences between before closure and after closure periods (p<0.01) and between closed season and after closure periods (p<0.01). Ordination of samples in a multi-dimensional scaling plot did not reflect any clear pattern and was not included.

SIMPER analysis for infaunal samples highlighted the traits "sexual maturity at less than 1 year" and "life span of less than 1 year" as the principal traits driving the differences between fished and control sites, both being more abundant in the control site. "Medium-large size" (more prevalent at the fished site) was also an important trait discriminating sites (Table 4).

SIMPER analysis performed only for infaunal fished samples revealed "high flexibility", "intermediate fragility" and "<1 y sexual maturity" as the main traits driving the differences between before closure and closed season periods, although they showed relatively low diss/sd (Table 4). All these traits were more abundant during the closed season. "Filter feeding" and "2 or more reproductive events per year", with relatively high diss/sd and also more abundant during the closed season, were important in discriminating between these two periods. Finally, "crawl" and "filter feeding" were the traits making the difference between the before closure and after closure period, both being more abundant in the latest (Table 4).

The SIMPER routine for epifaunal samples highlighted the traits "medium", ">5 y life span", "low flexibility" and "medium-large" as the most important traits driving the differences between control and fished sites, all of them being more abundant at the control site (Table 5). As the PERMANOVA test showed no significant interaction between site and time for epifaunal samples (Table 3), a SIMPER test based on the whole data set was performed to identify the traits driving the differences between effort regimes. This analysis highlighted "no flexibility", "deposit feeder", "permanent burrow dweller", "burrow" and "detritus"as the traits discriminating between the before and after closure periods and between the closed season and after closure periods. All these traits were more abundant in the before closure period and the closed season respectively. "Fragile", more abundant during the closed season, was also found to be important in discriminating between the closed season and the after closure period (Table 5).

Though these traits were highlighted by SIMPER analysis, it should be noted that average dissimilarities among fishing effort periods were low (10.03 and 8.65 for infauna and 16.33 and 13.83 for epifauna) (Tables 4, 5). Average dissimilarities between control and fished areas were also low (12.64 for infauna and 12.98 for epifauna).

Figures 5 and 6 show the trends of two traits for infauna (Fig. 5) and epifauna (Fig. 6) over the study period. These traits were highlighted by the SIMPER routine as being important in discriminating between fishing effort periods. Infaunal "filter feeding" and "two or more reproductive events" showed a decreasing trend over the study period at the control site but, whereas for "filter feeding" there seems to be no change at the fished site, in "two or more reproductive events" there was an increasing trend until the end of the closed season and afterwards this trait abundance decreased. Fishing effort values are overlapped in the figure, showing that epifaunal "deposit feeder" and "fragile" organisms were more abundant at fished and control sites when effort was low or null, in the autumn months. Plots for other traits are not included as they showed similar trends to those in Figures 5 and 6: "Filter feeder" trend for infauna was similar to "high flexibility" and highly similar to "<1 y of life span", and "<1 y sexual maturity". On the other hand, the "deposit feeder" trend for epifauna was highly similar to "detritus", "burrow" and "no flexibility", and the "fragile" trend was highly similar to "permanent burrow dwelling".

Table 6 shows that differences over fishing effort periods (time) for almost all these traits were significant, confirming that they were important in discriminating between these periods. Factor time was not significant for infaunal "filter feeder" and thus all the traits show a similar trend; however, the site:time interaction was significant, suggesting an influence of fishing effort periods or seasonality on these traits’ changes. Traits highlighted by SIMPER as the ones driving the differences between control and fished sites also showed significant differences for the factor site.

Different fish species have different habitats requirements, which could be more or less resilient to trawling impacts (Kaiser et al. 1999). The Sant Carles de la Ràpita fishing ground, our case study, is part of a nursery habitat for red mullet (Lombarte et al. 2000, Fiorentino et al. 2004), which is one of the most important commercial fish in the study area (Sánchez et al. 2007). Moreover, this species was chosen as a case study because it is a species closely related to the benthic system, with a well-known biology and life cycle which leads to particular catch dynamics (Fig. 4A). The existence of a nursery habitat is supported by the high catches of new recruits observed after the closed season, in September-October, which constitutes a specific characteristic of this species fisheries (Martin et al. 1999). Recruitment is a critical step for most fish life cycles and decline in recruitment may have important consequences for adult commercial stock (Bundy and Fanning 2005, Caddy in press). Therefore, the protection of benthic communities and habitat structure, which provide food and shelter for young fish, is essential to maintain a healthy adult stock.

As red mullet lives in a close relationship with the benthic environment for feeding, reproduction and refuge, this species might be particularly affected by the chronic alteration of benthic ecosystems (Auster and Langton 1998, Caddy in press). In this study we aimed to assess potential negative effects of trawling on red mullet due to changes on benthic functional components in the fishing ground (Fig. 7).

Regarding the food availability, small, short-lived infaunal organisms were more abundant at the control site, which might contribute to higher food production for red mullet in this site, especially for young recruits that feed on smaller prey (Machias and Labropoulou 2002). Moreover, during the closed season in summer, production of small infauna at the fished site may increase due to the higher prevalence of traits related to rapid reproduction (Fig. 7). The closed season coincides with the early phase of recruitment of red mullet (Fiorentino et al. 2008), so young red mullet recruits might benefit from higher food production in summer months. On the other hand, medium-large infaunal organisms dominated the fished site (Fig. 7), which could represent higher prey abundance for adult red mullet in the fishing ground. These results seems to contradict the current understanding of the effects of fishing on marine benthos, suggesting that communities in areas with a long history of fishing would consist mainly of small organisms (Jennings et al. 2001, Queirós et al. 2006). However, all medium-large taxa in the study area were deep burrowers that might avoid trawling disturbance (Brown et al. 2005), and also generally opportunistic-predator feeders that usually benefit from fishing activity (Frid et al. 2000).

Some of the observed trends in infaunal traits over the study period matched the fishing effort pattern, but the variability in infaunal abundance would more likely be related to seasonal patterns as the closed season in the fishing ground is too short to allow ecosystem recovery (Zajac 2003, Kaiser et al. 2006). The decreasing trend for almost all infaunal traits at the control site in summer and early autumn (closed and after closed season) could indicate a decrease in the overall infaunal abundance consistent with the temporal cycles shown by macroinfauna in the Mediterranean (Sardá et al. 1999). In contrast, the reduction of this seasonal trend at the fished site could be due to chronic fishing disturbance, which could alter the natural macroinfaunal cycles (de Juan et al. 2007b). However, this lack of trend in the fished site cannot be unequivocally linked to the fishing activities as natural processes could also play a role, resulting in an additive effect of both fishing and environmental variability (Koch et al. 2009).

In the epifaunal community, as expected and in agreement with de Juan et al. (2009), traits related to long life span and large size were more abundant at the control site (Table 5). This means that epifauna productivity (production/biomass) would be lower in the control area, although biomass production would be higher in this area. These large long-living organisms were fishes and large bivalves (e.g. Citharus linguatula or Acanthocardia echinata) that are not red mullet preys and, if they are predators, might compete with red mullet for prey. For example, Citharus linguatula, feeds in this area on decapoda and other small crustaceans (Juan et al. 2007a), which are potential prey for red mullet (Aguirre 2000), so these two fish partially share the trophic niche. Therefore, although adult red mullet could potentially find more food at the control site, this food would be less available due to interspecific competiveness (Fig. 7).

Regarding the habitat structure, as the control site is slightly deeper than the fished site, it might be a spawning area for red mullet. However, no traits related to habitat structure (e.g. sessile emerging epifauna) were highlighted in the analysis. In general, the whole fishing ground holds a homogenized community with a reduced habitat structure due to historical trawling disturbance in the area (de Juan et al. 2007b, 2009) (Fig. 7). Nevertheless, physical habitat is also important in creating habitat structure for recruits and spawners, and ROV images showed higher abundance of sediment structures such as ripples, mounds and pits at the control site (Demestre 2006). Changes in the trait "burrow" or "permanent burrow dwelling", which seemed to occur in similar abundance at fished and control sites over the sampled months, would not benefit red mullet spawning as these traits would destabilize the sediment with negative consequences for the creation of habitat structure (Lohrer et al. 2008). However, changes in these traits may affect bioturbation activity which may positively affect ecosystem production (Thrush and Dayton 2002, Lohrer et al. 2004) and indirectly benefit red mullet population in terms of food availability by enhancing primary production.

Associating demersal fish with their habitats is very critical to the definition of EFH and to correctly managing those EFH impacted by trawling activities (Kaiser et al. 1999). However, management strategies such as closed seasons are principally implemented to protect vulnerable steps of commercial species’ life cycles such as spawning and recruitment, focusing only on commercial species stock and not taking into account protection of benthic communities. Short-term closed seasons, such as the one implemented in the Sant Carles de la Ràpita fishing ground, do not enable benthic community recovery between successive periods of impact, especially on stable muddy bottoms where communities might take years to recover (Kaiser et al. 2006). Furthermore, short-term closures result in a concentration of the trawl effort and landings immediately after the closed season rather than in a more equitable pattern throughout the year (Martin et al. 1999). This concentration of fishing effort is also accompanied by an increase in discards (Sánchez et al. 2007), which indicates a higher level of disturbance on benthic communities. An adequate management regulation should progressively increase the fleet capacity after the closed season to avoid resource depletion and the highest disturbance levels on ecosystems.

Red mullet are an important commercial species in the Mediterranean, being one of the main target species for trawling fleets (Caddy 1993, Tserpes et al. 2002). In this case, the red mullet life cycle drives the fleet dynamics, as the highest fishing effort coincides with the highest landings (Fig. 4C), which are composed mainly of red mullet recruits (Demestre et al. 1997, Martín et al. 1999). Though the red mullet population could apparently tolerate this level of exploitation due to its high turnover rate, the yield per recruit shows evidence of overexploitation (Demestre et al. 1997) and this dominance of young specimens in landings makes the stock highly vulnerable to recruitment changes (Tserpes et al. 2002). Therefore, it would be advisable to protect red mullet nursery and spawning areas. Actually, Fiorentino et al. (2008) reported an increase in the number of recruits and a wider recruitment period after a trawl ban in the Gulf of Castellammare, exemplifying that an implementation of a permanently closed area does benefit reproduction success of this species. Moreover, a benthic community in a permanent closed area will have the possibility to recover from trawling impact, which might benefit not only red mullet but also other commercial species (e.g. the structured soft-bottom community observed in the Medes Island MPA, de Juan et al. 2011).

This work shows that changes in the effort regime within a year only had limited consequences for benthic community structure, whereas changes between a non-fished control site and a fished site were clearly evident. The observed changes at the fished site might benefit adult red mullet, as their food provision will be higher due to an increase in medium-large infauna and to lower interspecific trophic competition. However, red mullet recruits will be negatively affected by functional changes caused by fishing as their food provision might overall decrease, although they could benefit from a short-term increase in food production during summer. Moreover, both adults and recruits will suffer from lack of protection of habitat structures. Thus, the overall effect of trawling on the red mullet stock, considering the high fishing pressure on recruits and the indirect negative effects caused by ecosystem disturbance, could be a decrease in the spawning stock that will worsen the recruit's stock situation.

This study highlights the idea that permanent closure areas, which would allow recovery of the benthic ecosystem, restructuring habitats and communities, might be more beneficial for commercial species and their habitats than temporary closures.