Recruits identification

The first step in our analytical framework was the identification of recruits in the sampled population. Recruits were considered as those specimens that have settled on the bottom, becoming available to the fishing gear in well-defined habitats at the end of their larval—pelagic stage and which remain in these habitats before dispersing or migrating. Due to the lack of specific studies on the dispersal behaviour of juveniles (see for example [33]), we assumed that the first separable modal component of the age group 0 was composed of specimens recently settled on the bottom sharing similar habitat preferences. The Bhattacharya’s method [34] was used to separate the first modal component of the 0 age group from the annual standardized trawl survey length frequency distributions (LFDs). The 0 group threshold limit was defined using growth data available in the Mediterranean and routinely collected within the EU data collection framework (DCF). In some circumstances, when the splitting of modal components in the LFDs was difficult (e.g. low number of recruits), recruits were computed using a fixed threshold length derived from validated local studies. A summary of the data source, methods applied and threshold lengths used for recruits identification is provided in S1 Table.

The only exception to this framework of recruit identification was for common sole in GSA 17. For this population, the SoleMon survey sampled the 0 group juveniles when they migrate offshore from estuaries and coastal lagoons that are considered the core nursery habitats of the population [32]. Due to lack of data on small-sized sole in nursery areas, we addressed the position of open water juvenile habitats, where the largest specimens of the 0 group concentrated at the end of their inshore life. This information was considered crucial from a fishery management perspective to provide guidance for the protection of relevant juvenile habitats exposed to trawling activities.

As a second step in our framework we used the identified threshold length of recruits of each target species to calculate the density indices (n km^-2) of recruits in each sampling station as a base to model their annual spatial distribution.

Spatial modelling

From the results of the preliminary modelling trials carried out on a subset of species in several areas, it was clearly recognized that identification of a unique modelling approach/methodology suitable for all species/stages/areas was unrealistic. The amount of data (i.e. positive hauls), environmental characteristics of the areas and the availability of appropriate covariates were all factors that varied significantly across species and areas, in turn reducing the possibility to standardize the applied methods consistently. Spatial distribution modelling was also constrained by the number of positive hauls (i.e. hauls with occurrence of recruits of a given species) occurring for the different species and areas and proved unfeasible in many cases.

In our analytical framework we adopted the generalized linear model (GLM, [35]) as the main method for modelling the distribution of recruits. Model selection was performed with a forward stepwise selection procedure. Starting from the null model, each covariate was added individually; the predictor resulting in the highest AIC (Akaike Information Criterion) decrease or, depending on the model type, the better cross-validation index (CVI, S1 Text), was included in the model. The latter model was chosen to start the next step where the procedure was repeated until it was no longer possible to reduce AIC or improve CVI by including other available covariates. These included latitude, longitude, sampling depth, bottom steepness and distance from the coast. The latter two were obtained from the Marspec dataset [36].

When a spatial structure was evident in GLM residuals, different spatial correlation structures (e.g. exponential, spherical) were successively considered and the structure that best fitted the observations was identified [37]. This correlation structure was then incorporated as a random effect of a generalized linear mixed model (GLMM, [38]) where the final predictions were made by adding the predicted deterministic part of variation and the interpolated residual variation. Ordinary kriging (OK [37]), eventually replaced by the inverse distance weighting (IDW [39]) in case of poor fit of the variogram, was used to interpolate model residuals.

When the percent of total variance explained by a GLM/GLMM model was not satisfactory, we moved to other classes of models, first generalized additive models (GAM [40]) and then zero-inflated generalized additive models (ZIGAM [41]). In both cases we followed the same approach used for GLM to account for spatial autocorrelation in model residuals.

Finally, geostatistical methods, i.e. OK, and eventually IDW in case of poor variogram fit, were used when any of the regression models applied returned reliable spatial predictions.

Table 1 summarizes the modelling performed for recruits of the 11 selected species in the different Mediterranean sub-areas.

Generalized linear models (GLM) and generalized linear mixed models (GLMM)

GLM assumes that the dependent variable follows a distribution within the exponential family (e.g. normal, exponential, binomial) and a function of the mean (link function) depending linearly on some covariates.

The dependent variable in our case was the log of the density on site v and time t (y(v,t)) plus 1. The model used is: where the x_j(v,t)s are the covariates, β₀ is the intercept, β_j are the regression coefficients of the covariates.

When the variogram fit of the residuals showed spatial auto-correlation this was taken into account adopting a GLMM [38] where the final predictions were made by adding the predicted deterministic part of variation and the interpolated residual variation. Residuals were interpolated using ordinary kriging, eventually replaced by IDW, with the power coefficient generally fixed as 5, in case of poor variogram fit.

The general computation of this model is: where W(v,t) is the random effect that is assumed to be normally distributed.

In two specific case, recruits of the blackmouth catshark (G. melastomus) and deep-water rose shrimp (P. longirostris) in the Ligurian and North Tyrrhenian Seas (GSA 9), modelling issues raised by the high number of 0 observations were addressed by implementing a Bayesian GLMM with the INLA (Integrated Nested Laplace Approximations) approach [42]. The prior distributions of the parameters of the random effect were defined using an estimation chain where the posterior distributions of one year were used to define the prior distributions of the following year as in [25].

Generalized additive models (GAM) and generalized additive mixed models (GAMM)

The GAM is an extension of GLM, in which the linear form of the covariates function is replaced by a smooth function. The smooth functions used are generally cubic splines.

As for GLM, the dependent variable is log(y(v,t)+1) and the model was: where s(x_j(v,t)) is a non linear effect on the covariate.

Also in this case we checked for spatial autocorrelation in the residuals and, if needed, a kriging was accomplished to interpolate residuals variation to be incorporated into the model. In the latter case we have a GAMM with the following formulation: where, as in GLMM, W(v,t) is the random effect.

Zero inflated generalized additive models (ZIGAM)

One of the main problems in the analysis was dealing with the large amount of zero catches. ZIGAMs assume that the response variable follows a probabilistic mixture distribution of a zero atom and a continuous distribution belonging to the exponential family [41].

The data have been modelled in two steps. First the probability (P) of observing a density greater than 0 is modelled with a GAM where the dependent variable has a binomial distribution and the link function is the logit (logarithmic of the probability of observing a positive catch on the probability of observing a zero): Then a GAM model is estimated only on the positive catches using a Gaussian family model, the identity link function and log(y(v,t)) as dependent variable.

As in GLM and GAM we checked for spatial autocorrelation in the residuals and, if needed, an OK or IDW was used to interpolate residual variation to be incorporated into the model as random effects.

Kriging

Ordinary kriging was used to model the recruits of red mullet (GSA 18: Southern Adriatic Sea), common sole (GSA 17), common Pandora (GSA 10: South Tyrrhenian Sea; GSA 11: Sardinia, GSA 18), Norway lobster (GSA 11, GSA 19: Western Ionian Sea) and broadtail shortfin squid (GSA 17, Table 1).

The OK model structure can be formalised as: where μ(v) is the spatial trend (or large-scale variation).

Identification of annual nurseries and persistence analysis

As a third step of our approach to locate and classify nurseries, we identified the annual density hot-spots of recruits, which were regarded as annual nursery areas. The Getis’ G statistic [43], with a radius of 2.5–5.0 km and a 0.95 level of significance, was selected among the local methods for spatial hot-spot identification. This approach was applied separately in each GSA and for each year of the time series to spatially locate clusters of recruits displaying a significant higher density.

Finally, the Index of Persistence (I_i) measuring the relative persistence of the cell i of a given GSA as an annual nursery [25, 44] was calculated in each 1 km² cell considering δ_ij = 1 if the grid cell i was included in a nursery in the annual survey j, and δ_ij = 0otherwise: where n is the number of annual surveys considered in that GSA. The index I_i ranges between 0 (cell i never included in an annual nursery area) and 1 (cell i always included in an annual nursery area) for each cell of the GSA. Results were plotted in the persistence maps reporting five classes (0.05–0.20, 0.21–0.40, 0.41–0.60, 0.61–0.80 and 0.81–1). These maps allowed us to compare the nurseries persistence of the 11 target species at the scale of EU Mediterranean waters independently by the predicted recruit abundance.

Multispecies analysis of nursery area overlap

The spatial overlap between the identified nursery areas at different levels of temporal persistence (I_i >0.05, I_i >0.2; I_i >0.4) was examined to identify key recruitment areas in a multi-species context. In particular, the following levels were considered. In each GSA, every cell i was classified according to its importance for the recruitment of the set of species, considered using the following index of spatial overlap between recruitment areas: where k is the total number of modelled species showing I_ij > n^th-level in the GSA and δ(I_ij) = 1 if I_ij > n^th level in the cell i for the species j, and 0 otherwise. The index was applied in those GSAs where more than two species were modelled and classified each cell of a GSA between 0, no spatial overlap of persistent (I_i > n^th-level) nurseries and 1, complete spatial overlap.

Conservation effect of the current fisheries restricted areas (FRAs)

To evaluate the effect of the on-going temporal or permanent restrictions of trawl fishing activities in EU Mediterranean waters, we incorporated outcomes from the MEDISEH project. In particular, we used the data gathered from more than 200 trawl fishery restricted areas (TFRA) across the EU Mediterranean basins, where some sort of spatial (12 months per year) or temporal (>2 months per year) prohibition is applied (Fig. 2). TFRAs included trawl fishery restricted areas that may also have designations as marine protected areas under national legislation or have an international designation as TFRA by the EU/GFCM.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Trawl fishery restricted areas in the Mediterranean Sea.

Trawl fishery restricted areas are shown in blue, trawlable areas are shown in white.

https://doi.org/10.1371/journal.pone.0119590.g002

The spatial overlap between nurseries and TFRAs was calculated for each species to obtain the proportion of nursery areas currently under some sort of protection from trawling.

European hake

The threshold size of hake recruits ranged between 8.5 and 14.5 cm total length (TL). Nursery areas were mostly found between 100 and 250 m depth, with a patchy distribution along the shelf break (Fig. 3a).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Distribution maps of persistence index (Ii) of nursery areas of commercial fish in the North Mediterranean Sea.

A) European hake (Merluccius merluccius), B) red mullet (Mullus barbatus), C) common Pandora (Pagellus erythrinus), D) common sole (Solea solea), e) blackmouth catshark (Galeus melastomus) and E) thornback ray (Raja clavata). The index Ii ranges between 0 (portion i of the study area never included in an annual nursery) and 1 (portion i of the study area always included in an annual nursery).

https://doi.org/10.1371/journal.pone.0119590.g003

Along the Spanish coast the largest nurseries were identified between the Ebro River delta and Cape Nao between 150 and 250 m depth. Around the Balearic Islands, recruits were found at a low density north-west of the Menorca channel, and south-west of Mallorca. In the Gulf of Lions (GSA 7) recruits were found at the external border of the continental shelf, in the vicinity of the heads of canyons where a Fishery Restricted Area was established in 2008. A series of persistent relatively small but high density patches were found along the western coast of Italy (GSA 9: Ligurian and North Tyrrhenian Seas and GSA 10), on both sides of the Southern Adriatic Sea and Western Ionian Sea. Similarly, several areas of the Aegean Sea—Crete Island (GSA 22–23) and Eastern Ionian Sea (GSA 20) were characterized by high density areas of recruits with intermediate (I_i = 0.41–0.6) or high (I_i>0.6) temporal persistence, the most relevant being the one found in Saronikos Gulf. A less fragmented spatial pattern of nursery distribution was found in the Strait of Sicily (GSA 15–16), where the main nursery area was located to the east of the Adventure Bank extending also along the central-southern coast of Sicily, and off the south-western Sardinian coast (GSA 11).

Large persistent high density patches were also found in the central-northern Adriatic Sea (GSA 17) where three main persistent nursery areas were found respectively southwards and eastwards of the Middle Adriatic Pit and around the north-eastern margin of the South Adriatic Pit.

Red mullet

The timing of the MEDITS survey did not match the recruitment period of M. barbatus, as juveniles appeared mostly during late summer. However, in the Adriatic Sea (GSAs 17 and 18) it was possible to observe significant aggregation of recruits and consequently model their distribution (Fig. 3b). The threshold size of recruits in such areas ranged between 6.0 and 8.8 cm TL. In the Northern Adriatic a persistent recruitment area was identified between 10 and 20 m depth over the sandy-mud coastal areas of the Italian side. In the Southern Adriatic a main hot-spot of recruits was found offshore of the Gargano promontory, while a smaller one was observed in front of Bari (Apulia region).

Common Pandora

Persistent hot-spots of recruits of common Pandora were only successfully modelled along the coastal sandy areas of GSAs 17, 18, 10 and 11 (Fig. 3c). The threshold size of recruits averaged between 6.0 and 14.6 cm. Nursery areas were only evident in the central/northern Adriatic (GSA 17), with a large nursery spreading from the north-eastern Italian coast to the eastern coast of Istria (Croatia) at a depth of 5–20 m. Another important area was identified along the Southern Croatian coast. In the Southern Adriatic, South Tyrrhenian Sea and Sardinia, nursery areas generally appeared to be small and poorly persistent with a few exceptions (e.g. a small patch along the N Albanian coast).

Common sole

Common sole habitats could only be studied in GSA 17, where a time series of beam trawl survey data (SoleMon survey) was available. The threshold size of sole juveniles ranged between 17.7 and 20.4 cm TL. Their core distribution was located around the mouth of the Po River, whilst the southward extension of that area varied between years (Fig. 3d).

Blackmouth catshark

Nursery areas of G. melastomus were identified in all GSA areas, except in the waters surrounding Corsica (GSA 8), Cyprus, the Northern Adriatic, where the species is not present, and the Greek GSAs (GSAs 20–22–23, Fig. 3e). The assessment of threshold sizes for recruit identification varied among the areas, from a minimum value of 13.5 cm calculated in Greek waters, to a maximum of 28.5 cm TL applied in the Strait of Sicily. Persistent recruitment areas appeared as a long and almost continuous strip along the upper slope (200–600 m) off the western coast of Italy (GSAs 9 and 10), Gulf of Lions and the north-western coast of Sardinia (GSA 11). Along the Iberian Peninsula coastline (GSAs 1 and 6), most nursery areas showed a low temporal persistence except for one nursery located in the middle of the Alboran Sea and on the upper slope off the Balearic Islands, mainly to the north of Menorca Island and south-west of Mallorca.

Only scattered and poorly persistent nursery grounds were found in the north-western Ionian Sea, whereas on the upper slope of the neighbouring GSA 18, an area with a medium persistence (I_i = 0.4–0.6) appeared along the border of the South Adriatic Pit.

Three nursery areas with medium persistence were found on the upper slope of GSAs 15 and 16. The two largest ones were located to the north and south of the Malta Graben, and a smaller one was located to the south of Malta.

Thornback ray

Due to infrequent catches of thornback ray recruits, even in areas with rather high densities of adults (i.e. Corsica, Balearic Islands; [45]) modeling of the density data was feasible in only three GSAs (11, and 15–16). The threshold size of recruits ranged from 35–41 cm TL. Very few persistent nursery grounds were identified, with the more important areas located in the Strait of Sicily (GSAs 15–16) at depths up to 400 m (Fig. 3f). The most stable density patch occurred offshore from the south-western coast of Sicily, in an area of intense hydrographic activity due to the eastward surface flow of Modified Atlantic Water and the westward flow of Levantine Intermediate Water. High density patches of recruits were also found around Sardinia, at depths of 100–300 m (Fig. 3f).

Giant red shrimp

Giant red shrimp nurseries were identified in the central Mediterranean Sea, including South Tyrrhenian Sea, Sardinia, Strait of Sicily, Southern Adriatic Sea and Western Ionian Sea, Fig. 4a. MEDITS catches from the Western Mediterranean (GSAs 1, 5–7), the Northern Adriatic and Cyprus were extremely low. Around Corsica Island, in the Ligurian and North Tyrrhenian Sea, as well as in the Eastern Ionian, Aegean Sea and Crete Island, giant red shrimp where somewhat more abundant, but nevertheless the catch of recruits was insufficient for nursery grounds identification.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Distribution maps of persistence index of nursery areas of commercial species of crustaceans and cephalopods in the North Mediterranean Sea.

A) Giant red shrimp (Aristeomorpha foliacea), B) deep-water rose shrimp (Parapenaeus longirostris), C) Norway lobster (Nephrops norvegicus), D) horned octopus (Eledone cirrhosa), E) broadtail shortfin squid (Illex coindetii). The index Ii ranges between 0 (portion i of the study area never included in an annual nursery) and 1 (portion i of the study area always included in an annual nursery).

https://doi.org/10.1371/journal.pone.0119590.g004

The average threshold size of recruits in the areas varied between 25 mm and 31 mm carapace length (CL). The depth range of the most persistent density hot-spots generally varied between 500 m and 700 m depth, although a few nurseries were found to be located in shallower waters on the upper continental slope starting at depths of 250 m.

In the South Tyrrhenian Sea (GSA 10) nursery areas with an intermediate level of temporal persistence (I_i = 0.4–0.6) were localized off the Calabrian coast. Around Sardinia, nursery areas were found both on the east and west side, with the most persistent patch of recruits offshore of the south coast. In the Strait of Sicily (GSA 15–16), recruits were abundant in a wide area with an intermediate to high level of persistence (I_i = 0.4–0.8) off the south to south-west coast of Malta. Two persistent patches of recruits were found south of the Otranto channel (GSA 19) and off the southern border of the Adriatic Pit (GSA 18).

Deep-water rose shrimp

Most GSAs in the Mediterranean were characterized by high abundances of recruits of P. longirostris. Exceptions were the Balearic Islands, the Gulf of Lions, Corsica, the Eastern Ionian and Cyprus, where the few positive hauls in each annual survey and/or the low amount of individuals captured did not allow the distribution of nursery areas to be modelled. Overall, mean length of recruits ranged between 10 and 20 mm CL. Nursery areas were mostly located on muddy bottoms along the shelf break-upper slope between 100–200 m and even deeper in some areas (Fig. 4b). A single wide nursery area occurred all along the South Spanish coast (mainly in GSA 1), with the highest persistence (I_i = 0.6–0.8) found in the eastern North Alboran Sea. Off the Italian coast nurseries appeared as highly persistent (I_i = 0.8–1.0) patches, in particular in the Tyrrhenian Sea (GSA 9 and 10), western coast of Sardinia, Strait of Sicily and Western Ionian Sea. Three large areas of high persistence (I_i = 0.8–1.0) were identified for the recruits of deep-water rose shrimp in the Northern Adriatic. They extend south east of the Middle Adriatic Pit until the southern limit of GSA 17. In the Aegean Sea and Crete Island the most persistent density hot-spots (I_i = 0.4–0.6) were identified in offshore deep areas of Argolikos Gulf (west-central Aegean), in the Thracian Sea (northern GSA 22) and on the northern Cretan shelf (GSA 23).

Norway lobster

Norway lobster recruits were below 20–30 mm CL, depending on the area. Nursery areas were found at depths ranging from 100 to 600 m on the north-eastern coast of Sardinia, in the Strait of Sicily, the Northern Adriatic Sea and along the south-eastern coast of Italy (north-western Ionian Sea, Fig. 4c). Very low MEDITS catches of juveniles were obtained in the remaining GSAs. Along the upper slope of Sardinia nursery areas were identified as thin and relatively poorly persistent patches at depths ranging from 400 to 600 m. In the Strait of Sicily recruits were distributed on muddy bottoms of the upper slope between 250 and 500 m. Two major large offshore nurseries were identified along the south-eastern coast of Sicily. Another large and persistent (I_i = 0.4–0.6) offshore recruitment area was found in the Western Ionian Sea on the Apulian coast (Italy).

In the Northern Adriatic Sea a wide area where small-specimens concentrate was identified in correspondence with the Middle Adriatic Pit at depths ranging from 100 m to 270 m.

Horned octopus

Recruits were identified using a threshold size between 30 and 56 mm mantle length (ML). Along the Spanish coast and the Strait of Sicily MEDITS catches of juveniles were too low to develop reliable spatial modelling. Modelling showed the occurrence of several small and persistent density hot-spot areas along the shelf-break off the north-western coast of Italy (GSA 9) and in Sardinian waters, mainly over detritic bottoms colonized by crinoid beds between 100 and 180 m depth (Fig. 4d). In the central-southern Tyrrhenian Sea (GSA 10) two very persistent (up to 100%) hot-spots were localised in the Gulf of Gaeta and Napoli.

Nurseries appeared as large and highly persistent patches from the shelf-break to upper slope of the Gulf of Lions and in the Northern Adriatic. Here a wide offshore recruitment area was identified between 60–160 m depth on the south-western side of the GSA. This area appeared connected with a series of small patches of recruits on the margin of the Adriatic Pit (GSA 18). A second, smaller area was found westwards of the Croatian Islands at a depth of 120–150 m (Fig. 4f). In the Greek Seas (GSAs 20, 22–23) the main nurseries were found in the north-western part of the Aegean Sea, mainly on the extended continental shelf of Thermaikos Gulf (Fig. 4d).

Broadtail shortfin squid

Nurseries of the broadtail shortfin squid were identified in most of the GSAs using a threshold size of recruits between 61 and 115 mm ML. Along the Spanish coast recruits were more abundant in the Catalan Sea (GSA 06), with the highest abundance observed between 100 and 200 m depth. In the Ligurian and North Tyrrhenian Sea recruits were distributed in several small patches along the shelf break mainly between 70 and 150 m depth (Fig. 4e). Two main, rather persistent (I_i = 0.4–0.6), nursery areas appeared offshore in the central and southern sectors of the western Sardinian coast, between depths of 100 and 300 m, with an increased abundance over the detritic bottoms at the shelf-break. Similarly, in GSA 10, persistent nursery areas (I_i = 0.4–0.6) were localized on detritic bottoms in the Gulf of Gaeta and off Capo Bonifati between 70 and 150 m depth (Fig. 4e). In the Strait of Sicily, the main nursery areas were located offshore off the south-east edge of the Adventure Bank (South-East Sicily). Another important nursery occurred along the eastern edge of the Malta Bank (Fig. 4e). In the Northern Adriatic there was a wide recruitment area extending between 50 and 300 m depth with two more persistent patches. Nurseries in the Southern Adriatic Sea (GSA 18) mainly occurred along the Italian side, between 100 and 150 m depth. Small and persistent high density patches of recruits were also identified in the Ionian Sea both on the Italian (GSA 19) and Greek (GSA 20) sides, with the most persistent areas being outside Thermaikos Gulf in the Northern Aegean (GSA 22).

Multispecific spatial overlap between nursery areas

The analysis of the spatial overlap of the nurseries identified at different persistence levels (I_i >0.05, I_i >0.2 and I_i >0.4) allowed the identification of several areas across the Mediterranean that provided multispecies critical habitats, and hence the exploration of different scenarios (Fig. 5). No overlap analysis was performed in GSA 5 where only two species were modelled.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Map showing the spatial overlap among the persistent nursery areas of 11 species of demersal fish and shellfish in EU Mediterranean waters.

Colours indicate different classes of spatial overlap index (SO). Spatial overlap was calculated considering three different classes of the temporal persistence index (I_i): A) I_i>0.05; B) I_i>0.2; C) I_i>0.4. Numbers indicate the numbers of species modelled in each GSA. Note that, since the number of species modelled varied among GSAs, the overlap index SO was calculated separately for each GSA to avoid any comparative analysis at a larger geographical scale that can lead to misleading conclusions.

https://doi.org/10.1371/journal.pone.0119590.g005

It is important to note that, since the number of species modelled varied among GSAs, the overlap index SO_i can be analysed only at the GSA level, avoiding any comparative analysis of its values at a larger geographical scale that can lead to misleading conclusions. However, the emergence of some recurrent, and even important, patterns of overlap can be highlighted.

As expected, the overlap areas decreased drastically when increasing the level of persistence from I_i >0.05 (Fig. 5a) to I_i >0.4 (Fig. 5c). Several patches of co-occurrence of nursery areas of multiple species were identified for I_i >0.4 only along the coast of the Ligurian and North Tyrrhenian Sea, Northern Adriatic Sea, Strait of Sicily, Aegean Sea and Crete Island (Fig. 5c). The lower interspecific spatial overlap observed in the other GSAs can be attributed to both a low level of persistence of many of the species and a considerable spatial mismatch among the most persistent nursery areas of the different species.

Nursery areas of European hake, deep-water rose shrimp, broadtail shortfin squid and horned octopus co-occurred along the shelf-break with spatial difference in temporal stability, from I_i >0.2 off the south-west Sardinia and the northern sector of the south Tyrrhenian (Fig. 5b) to I_i >0.4 in the central-northern Tyrrhenian Sea and central Adriatic (Fig. 5c). In the other GSAs, the nursery areas of these 4 species, when identified, also showed a clear spatial association. The overlap area was due to the co-occurrence of European hake and broadtail shortfin squid along the Spanish coast and in the Eastern Ionian for Ii >0.2, also associated with the deep-water rose shrimp in the Strait of Sicily and along the South East Sicilian coast (GSA 19) and with the blackmouth catshark in the Gulf of Lions and in the northern sector of GSA 10. The nurseries of European hake were found in partial overlap with the nurseries of the two cephalopods in the North-West side of GSA 18 while on the east side of the GSA, along the Albanian coast, the species prevalently co-occurred with the deep-water rose shrimp. Varying degrees of overlap between nurseries areas of European hake and deep-water rose shrimp in association with the horned octopus were also found in the Aegean Sea and Crete and in North Sicily (GSA 10). A degree of overlap between the distribution of broadtail shortfin squid and deep-water rose shrimp recruits was observed in the Northern Alboran Sea (GSA 1) independently of persistence level (Ii >0.05, Fig. 5a). Overlap of stable nurseries of the more coastal species, red mullet, common Pandora and common sole were exclusively observed in the Italian waters of the Northern Adriatic. Finally, with regard to bathyal species, patches of co-occurrence of giant red shrimp and Norway lobster were identified in the Western Ionian and in the deep basin between the two banks of the Strait of Sicily, sometimes in association with the blackmouth catshark. A nursery area of blackmouth catshark partially overlapped a nursery area of giant red shrimp on the deep grounds of the northern sector of GSA 10.

Proportion of nurseries under fisheries restrictions

The on-going spatial measures of fisheries restrictions for trawl fisheries (TFRA) in the Mediterranean Sea showed a significant conservation effect mostly for species with nurseries on coastal grounds shallower than 50 m. This was the case of red mullet, common Pandora and common sole with 66.8%, 54.1% and 46.1% of persistent nursery areas under protection (Table 2). On the other hand the proportion of nursery areas protected was much lower for European hake (15.3%) and Norway lobster (17.8%), whose recruitment occurs on the shelf-break and upper slope below 150 m. Finally, there was minimal overlap between TFRA and recruitment areas for offshore species (i.e. thornback ray, blackmouth catshark, giant red shrimp and deep-water rose shrimp) distributed over the continental slope deeper than 200 m.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Surface extent in km2 of total nursery areas identified and of nursery areas with a temporal persistence >40% (N40), for a set of 11 demersal species in Mediterranean European waters.

https://doi.org/10.1371/journal.pone.0119590.t002