Deep-water fish assemblages in the Mediterranean Sea *

Data collected during the DESEAS survey carried out in three areas of the Mediterranean Sea (Balearic Sea, western and eastern Ionian Sea) were used to describe the structure and spatial distribution of the fish assemblages at depths between 600 and 4000 m. A total of 55 species (8 chondrichthyes and 47 teleost fish) were sampled, 38 of them in the Balearic Sea, 30 in the western Ionian and 37 in the eastern Ionian. Multivariate analysis showed a clear pattern of zonation with depth and geographic area. Three main assemblages were identified across the vertical gradient investigated: on the upper slope around a depth of 600 m, on the middle slope between 800 and 1300 m, and on the lower slope below 1300 m. The geographic characterisation of the ichthyofauna structure was mainly observed on the upper and middle slope. The highest abundance and biomass values were found at depths of 1000-1200 m due to the presence of larger species with high energy requirements, such as Alepocephalus rostratus, Mora moro and Galeus melastomus. On the middle slope, significant differences in the abundance, biomass and mean fish weight were detected between the three areas. Significant differences for these parameters were also shown between the middle and lower slope assemblages. A significant decrease in species richness and diversity was shown with increasing depth, indicating an impoverishment of the megafauna with depth and a higher similarity between the Balearic Sea and the Ionian Sea with regard to the greatest depths. The dominant species on the deepest bottoms were Bathypterois mediterraneus, Chalinura mediterranea and Coryphaenoides guentheri. On the lower slope, the shark Centroscymnus coelolepis was also found to be relatively abundant off the Balearic Islands, Etmopterus spinax in the eastern Ionian Sea and the teleost fish Cataetyx laticeps throughout the three study areas. The deepest bottoms of the Mediterranean Sea shelter an ichthyofauna dominated by small-medium species living in a food scarce environment in which some large mobile fishes are widespread. The results are discussed in relation to previous studies mostly carried out in the Atlantic and western Mediterranean.

SUMMARY: Data collected during the DESEAS survey carried out in three areas of the Mediterranean Sea (Balearic Sea, western and eastern Ionian Sea) were used to describe the structure and spatial distribution of the fish assemblages at depths between 600 and 4000 m. A total of 55 species (8 chondrichthyes and 47 teleost fish) were sampled, 38 of them in the Balearic Sea, 30 in the western Ionian and 37 in the eastern Ionian. Multivariate analysis showed a clear pattern of zonation with depth and geographic area. Three main assemblages were identified across the vertical gradient investigated: on the upper slope around a depth of 600 m, on the middle slope between 800 and 1300 m, and on the lower slope below 1300 m. The geographic characterisation of the ichthyofauna structure was mainly observed on the upper and middle slope. The highest abundance and biomass values were found at depths of 1000-1200 m due to the presence of larger species with high energy requirements, such as Alepocephalus rostratus, Mora moro and Galeus melastomus. On the middle slope, significant differences in the abundance, biomass and mean fish weight were detected between the three areas. Significant differences for these parameters were also shown between the middle and lower slope assemblages. A significant decrease in species richness and diversity was shown with increasing depth, indicating an impoverishment of the megafauna with depth and a higher similarity between the Balearic Sea and the Ionian Sea with regard to the greatest depths. The dominant species on the deepest bottoms were Bathypterois mediterraneus, Chalinura mediterranea and Coryphaenoides guentheri. On the lower slope, the shark Centroscymnus coelolepis was also found to be relatively abundant off the Balearic Islands, Etmopterus spinax in the eastern Ionian Sea and the teleost fish Cataetyx laticeps throughout the three study areas. The deepest bottoms of the Mediterranean Sea shelter an ichthyofauna dominated by small-medium species living in a food scarce environment in which some large mobile fishes are widespread. The results are discussed in relation to previous studies mostly carried out in the Atlantic and western Mediterranean.

INTRODUCTION
The various oceanographic expeditions carried out over the last two centuries, among which the Challenger was the most famous (Ryland, 2000), mostly increased knowledge on the taxonomy and biogeography of the fish species collected in the various geographic areas and at different depths. Studies outlining the ecology and zonation of deep-sea ichthyofauna only started in the 1960s and most of the progress in this direction has been achieved in both the western and eastern north Atlantic as a result of commercial and scientific trawling surveys, sampling with traps and long lines, and observations from submersibles and baited cameras (e.g. Isaacs and Schwartzlose, 1975;Haedrich and Rowe, 1977;Gordon, 1979;Hureau et al., 1979;Haedrich et al., 1980;Merrett and Marshall, 1981;Gordon and Duncan, 1985;Haedrich and Merrett, 1988;Hecker, 1990;Gage and Tyler, 1991;Merrett et al., 1991;Gordon and Bergstad, 1992;Merrett and Haedrich, 1997).
These studies have provided many insights into the life adaptations and strategies, diversity and distribution of deep benthic and benthopelagic fish as well as vertical patterns and zonation of the fish assemblages. For both diversity and vertical zonation, investigated down to the abyssal plane, their patterns have been shown to vary from area to area in the north Atlantic (Snelgrove and Haedrich, 1985;Haedrich and Merrett, 1990;Hecker, 1990;Koslow, 1993). The sampling limitations and the different selectivity of each gear used can play an important role when the fish assemblages are compared between different areas. In other words, the assessment of biodiversity in the deep sea is highly influenced by the gear used. However, the pattern of diversity and vertical zonation are well known to be determined by both geomorphologic and oceanographic factors, such as depth, substrate type and hydrography, and biological conditions, such as resource availability, predator-prey relationships and intraspecific and interspecific competition (Haedrich et al., 1980;Merrett, 1987;Mahaut et al., 1990;Bianchi, 1992;Koslow, 1993).
The DESEAS project, funded by the EU, gave the opportunity to investigate the deep-sea fish fauna in three areas of the Mediterranean Sea for the first time, during the same period and with the same vessel and gear. Despite the several limitations of sampling bathyal and abyssal fish species (Merrett and Haedrich, 1997), this study represents the first attempt to identify deep-sea fish assemblages across a depth range of 600-4000 m and along a geographic gradient from the western to eastern Mediterranean.
The sampling was conducted with the otter trawl Maireta System (OTMS) using the R/V García del Cid (38 m long; 1500 HP). The OTMS (height 1.8-2 m; horizontal opening 14 m) was trawled by a single warp and operated with a pair of rectangular iron otter boards (1.20 x 2.0 m; 450 kg). A detailed description of the net can be found in Sardà et al. (1998). The initial and final tow time was measured by means of the SCANMAR sonar system. The ves-sel speed and geographical position were measured using differential GPS. Profiles of turbidity, temperature and salinity were taken using a CTD probe from the surface to the bottom in each study area. Trawling was carried out during daylight hours. The number of hauls carried out in each area, with the indication of the depth, is reported in Table 1.
During the DESEAS survey, despite the long time employed in the exploration of trawlable bottoms, the sampling design was not completely satisfactory due to the different bathymetry and topography of the three areas. Many zones, mostly at depths greater than 1500 m, prevented trawling or caused the breaking of the net due to irregular bottoms. This explains the unbalanced number of hauls between the three areas, mostly below the abovementioned depth.
Specimens were identified on board following the nomenclature reported in Whitehead et al. (1984Whitehead et al. ( , 1986aWhitehead et al. ( , 1986b. Haul duration ranged from 1 to 3 hours. The results obtained in number and weight were standardised to km 2 for subsequent numerical processing. Basic data were log-transformed to reduce the influence of the most abundant species. Classification and ordination were performed in order to identify demersal faunal assemblages. The former was carried out by means of cluster analysis using the Euclidean distance coefficient and the complete linkage and UPGMA methods (Ludwig and Reynolds, 1988), the latter through non-parametric multi-dimensional scaling (MDS) (Kruskall, 1964 Species recorded in a single haul and those with pelagic habits were excluded from the computations. Only Lampanyctus crocodilus was included in the analysis since the adult specimens of this species live close to the bottom . All these analyses were carried out using STATISTICA software (StatSoft, 1995).
Considering the different difficulties of this kind of research (among which the long time required for the exploration of unknown deep-sea bottoms suitable to be trawled, the long time required to make a single haul and the decreasing sampling efficacy with depth), the term "assemblage" refers to a cluster of fish species available to the used gear in the investigated areas irrespective of whether they interact or not (Wootton, 1991), thus avoiding any confusion with other ecological meanings such as "community". The average abundance (N Km -2 ), biomass (g Km -2 ), mean fish weight (g/fish), species richness and Shannon-Wiener diversity index (H') were determined for the main groups resulting from the classification and ordination. The geometric mean was adopted in computation of abundance, biomass and mean fish weight to minimise the negative effect caused by extreme values (Stefanescu et al., 1993). Differences between groups were evaluated using the non-parametric test of Kruskal-Wallis (Conover, 1980). Changes in the abundance, biomass, mean fish weight, species richness and diversity index with depth were evaluated both with regression analysis and Spearman non parametric correlation (Conover, 1980).

RESULTS
A total of 55 species (8 chondrichthyes and 47 teleost fish) were collected during the DESEAS survey and are listed in Table 2.
Excluding the mesopelagic and bathypelagic species (with the exception of L. crocodilus), 32, 24 and 31 species were captured off the southern Balearic Islands and in the western and eastern Ionian Sea respectively.
The resulting dendrogram of hauls indicates the presence of four main clusters related to the depth and geographic area (Fig. 2). The first cluster consisted of the two hauls from depths of 1000 and 1200 m in the Balearic Sea. The second cluster concerned the deepest stations covering the depth range between 1500 and 4000 m in all three areas. A third group was made up of the stations investigated between 800 and 1500 m in the Ionian Sea, both western and eastern. The fourth aggregation consisted of the shallower hauls carried out around 600 m together with the station conducted at 802 m in the Balearic Sea. In this last cluster a lower distance (higher similarity) was shown between the two hauls carried out at 600 m in Greek waters, due to the remarkable abundance of Chlorophthalmus agassizii and Helicolenus dactylopterus.
The two hauls at depths of 1000 and 1200 m in the Balearic Sea showed a higher similarity (lower distance) in relation to the exclusive presence of the same depth showed the lowest distance values. Within the cluster of the deepest hauls, the two deepest hauls (3300 and 4000 m) were grouped together with one haul carried out at 1500 m in the same area. This latter sub-group is the result of a very low number of species at all three stations with the common occurrence of Bathypterois mediterraneus found in comparable densities.
The results of MDS (stress level = 0.133) are shown as a two-dimensional representation in Figure 3. The ordination of the 26 hauls was in general agreement with the clustering, showing the aggregation of the shallower and deeper hauls according to the positive and negative values of the first dimension respectively. In fact, the first dimension of the axes was highly correlated with depth (r spearman = -0.864; P < 0.01) (Fig. 4). Moreover, the MDS representation shows more clearly that the hauls at 1500 m are grouped together with those carried out on the deepest bottoms.
A greater heterogeneity was evident in the ordination of the stations carried out between 800 and 1300 m. These stations reflect the geographic location due to the different abundance of the species in the three areas. The western Ionian stations are diagonally positioned between the Balearic and eastern Ionian ones according to the second dimension of the axes.
According to the results of the multivariate analysis, the values of the average abundance, bio-mass, mean fish weight, mean species richness and diversity index were computed for the groups of hauls carried out at depths of 600-650 m (upper slope), 800-1300 m (middle slope) and below 1300 m (lower slope). The ecological parameters in the three areas were only calculated on the middle slope where a sufficient number of samples were collected at the same depths (Table 3).
Comparing the results between the three areas using the Kruskal-Wallis test, no significant differences were shown for species richness and diversity. Concerning the abundance, significant differences were detected between the western Ionian and eastern Ionian (P = 0.02) and between the Balearic Sea 92 G. D'ONGHIA et al. and eastern Ionian (P = 0.006), while the differences were not significant between the Balearic Sea and western Ionian Sea. With regard to biomass, significant differences were shown between the Balearic Sea and western Ionian Sea (P = 0.05) and between the Balearic Sea and eastern Ionian (P = 0.0007), while no significant differences were observed between the western and eastern Ionian.
The mean fish weight values showed significant differences between the three areas (P = 0.028 for Balearic-western Ionian; P = 0.0005 for western Ionian-eastern Ionian; P = 0.0001 for Balearic-eastern Ionian).  Comparing the ecological parameters of the middle slope as a whole (800-1300 m) with those of the lower slope (> 1300 m), the Kruskal-Wallis test gave significant differences for the abundance (P = 0.043), biomass (P = 0.0017) and mean fish weight (P = 0.0001).
The percentages of abundance and biomass of the dominant species across the depth range are reported in Table 4. In this table, the percentage values in the three areas were only calculated in the middle slope for the abovementioned reasons; the lower slope assemblage was further divided into more coherent sub-groups according to depth.
In the upper slope group (600-650 m), though H. dactylopterus was found almost exclusively in Greek waters, it was the dominant species both in biomass and abundance. C. agassizii, H. mediterraneus and Phycis blennoides were among the most abundant species both in number and weight. The former was exclusively found in the eastern Ionian Sea.
Concerning the middle slope (800-1300 m), in the Baleric Sea the dominant species both in number and weight was A. rostratus. Lepidion lepidion, N. aequalis and Galeus melastomus provided noteworthy contributions to the abundance and Mora moro and G. melastomus to the biomass. In the western Ionian Sea the most abundant species was M. moro followed by Nezumia sclerorhynchus and G. melastomus. The greatest biomass contributions were made by M. moro, G. melastomus and Dalatias licha. In the eastern Ionian Sea the most abundant species were G. melastomus, M. moro and L. crocodilus in density and Hexanchus griseus, M. moro and G. melastomus in biomass.
On the lower slope, at 1500-2000 m, B. mediterraneus was the dominant species in number. It was mostly found in the Ionian Sea (54% on the western side, 11% on the eastern and 4% in the Balearic Sea). Although A. rostratus was exclusively captured off the southern Balearic Islands down to 1500 m, it provided the greatest biomass contribution in this depth stratum.
At depths of 2200-2800 m, B. mediterraneus was the most abundant species followed by Chalinura mediterranea, Coryphaenoides guentheri and Cataetyx laticeps. The dominant species in biomass was Centroscymnus coelolepis followed by C. laticeps.
On the abyssal plane (3300-4000 m) the dominant species was C. mediterranea both in number and weight.

DISCUSSION
In spite of the sampling limitations encountered throughout the three investigated areas, this study reveals the existence of deep water fish assemblages in which depth and geographic location play the main role in megafaunal group differentiation.
Basically, the faunal zonation by depth can be identified in an upper slope (around 600 m), middle slope (800-1300 m) and lower slope (> 1300 m) assemblage as shown in the north Atlantic (Haedrich et al., 1980;Haedrich and Merrett, 1988;Hecker, 1990) and in the western Mediterranean (Stefanescu et al., 1993;Moranta et al., 1998). In agreement with these studies, the peak of abundance and biomass was found around 1200 m and the main faunal change along the depth gradient was shown at around 1500 m. According to Hecker (1990) and Stefanescu et al. (1993), this bathymetric zone can be considered to be within a "transitional" zone where an important species turnover occurs. The former author found this transitional zone varying from a depth interval ranging from 100 m (1200-1300 m) to 400 m (1200-1600 m) in four geographic locations in the north Atlantic, reflecting differences in local topography and geology. The latter identified a transitional zone in the lower slope around 1400 m due to the substitution of some dominant and subdominant species. Cartes and Sardà (1992) and Cartes (1993), studying the deep-sea decapod crustaceans, reported a faunal discontinuity at around 1200 m in the Catalan Sea.
The small differences in the boundaries between the middle and lower slope shown in the various studies (op. cit.) might be due to the different number of samples and to the geographic characterisation of the trawl stations (Haedrich and Merrett, 1990;Koslow, 1993). This latter aspect was particularly evident in the upper and middle slope, where the main differences in the faunistic abundance between the Balearic and Ionian areas were detect-ed. Although the small number of hauls and the gear selectivity could affect the finding of the different species whose abundance appears to decrease with depth, the abundant presence of C. agassizii and H. dactylopterus exclusively on the upper slope of the eastern Ionian Sea might be due to the absence of fishing activity in this area below 400-500 m in depth (D'Onghia et al., 2003). Though on the deepest bottoms the occurrence of some species and their abundance (e.g. A. rostratus only found off the Balearic Islands down to 1500 m) influenced the linkage between the hauls, the assemblage detected on the lower slope indicates a higher homogeneity between the three areas and thus a higher similarity in the ecological conditions between the Balearic Sea and the Ionian Sea at the greatest depths.
Considering the stability of the environmental parameters in the deep water of the Mediterranean Sea, the ichthyofauna zonation has been attributed to the trophic resources available above and below a depth of 1200 m linked to the distribution of mesopelagic organisms and the particular hydrological conditions of this basin (Carrassón and Matallanas, 1990;Stefanescu et al., 1992bStefanescu et al., , 1993Morales-Nin et al., 1996;Moranta et al., 1998;Maynou and Cartes, 2000;Carrassón and Matallanas, 2001). The larger species with high energy requirements, such as A. rostratus, Trachyrhynchus trachyrhynchus, M. moro and G. melastomus are replaced at greater depths by smaller species which utilise a broad spectrum of benthopelagic prey, such as B. mediterraneus, C. mediterranea and C. guentheri. However, on the lower slope the presence of large fishes, such as C. coelolepis off the Balearic Islands, Etmopterus spinax in the eastern Ionian Sea and C. laticeps in all three study areas, is also important in the structure of the deep-sea Mediterranean ichthyofauna.
Although it is generally accepted that the megafauna changes between the middle and the lower slope may be due to the sharp reduction in the mesopelagic fauna, its excretory products and detritus sedimenting down from shallower layers, most of the studies on these bioenergetic aspects have been conducted in the Atlantic (e.g. Gordon, 1979;Vinogradov and Tseitlin, 1983;Merrett, 1987;Gage and Tyler, 1991;Mauchline and Gordon, 1991), while very little is known about them in the Mediterranean (Stefanescu et al., 1993) and particularly in the eastern basin.
In the context of the oligotrophic Mediterranean conditions, abundance, biomass and mean fish weight values came out significantly greater in the Balearic Sea than in the western and eastern Ionian. In addition, the abundant occurrence of A. rostratus (between 800 and 1500 m) and C. coelolepis (on the lower slope) was only recorded in the Balearic Sea. These differences between the study areas could be linked to the fact that the western basin shows a higher productivity than the eastern one (Danovaro et al., 1999;Maynou and Cartes, 2000;Melley et al., 2000). In fact, megafaunal biomass in the deep sea largely depends on the surface productivity regime (Haedrich and Rowe, 1977;Gordon 1979;Merrett, 1987). A. rostratus feeds mainly on gelatinous macroplankton (Carrassón and Matallanas, 1990) and it has never been recorded in the eastern Mediterranean. In the Atlantic, Anderson et al. (1985) and Merrett (1987) reported that the Alepocephalidae family is abundant in oligotrophic and less seasonal areas and is usually replaced by other families, such as Macrouridae and Moridae, in more eutrophic areas. This contrasts with our findings, since Macrouridae and Moridae were captured throughout the basin compared to A. rostratus only on the western side, which is assumed to be less oligotrophic than the eastern side. In addition, Moranta et al. (1998) reported that A. rostratus is more abundant south of the Balearic Islands (Balearic Sea in this study) than to the north of these islands, where the species L. lepidion and B. mediterraneus are more abundant. These differences seem to be due to the differences in the trophic web (Massutí et al., 2004). South of the Balearic Islands, where the food of planktonic origin is more important, the dominant species is A. rostratus, which feeds mainly on macroplankton (Carrassón and Matallanas, 1990), while north of the Balearic Islands, where there are submarine canyons and the benthic and benthopelagic organisms play an important role in the food web (Cartes et al., 1994), the most abundant species are L. lepidion and B. mediterraneus, which prey on benthic decapods (Carrassón et al., 1997) and benthopelagic crustaceans (Carrassón and Matallanas, 2001) respectively. These different conditions in the trophic web, due to the surface productivity and the presence of canyons, could explain the absence of A. rostratus in the Ionian Sea, where submarine canyons on the slope are widespread (D'Onghia et al., 1998b). Reflecting these remarks, A. rostratus and B. mediterraneus showed striking dominance in the Balearic Sea and Ionian Sea respectively.
The reason for the absence of C. coelolepis in the samples taken in the Ionian Sea during this study remains unclear. In fact, the presence of C. coelolepis is known in the eastern Mediterranean. It has been recorded together with H. griseus, G. melastomus, E. spinax and Centrophorus spp. between 1500 and 2500 m in the Cretan Sea, and between 2300 and 3850 m in the Rhodes Basin (Jones in Priede and Bagley, 2000;Jones et al., 2003). However, the presence of shark species recorded during this and previous studies in both the western and eastern Mediterranean Sea indicates that the trophic system depending on the pelagic input and involving large mobile fishes occurs throughout the basin, although with possible qualitative-quantitative differences between the western and eastern side.
Compared with the Atlantic, the depths of the Mediterranean show an impoverishment of megafauna as well as lower abundance and biomass in the structure of the ichthyofauna. These differences have long been considered to be due to historical factors, energetic constraints and physical conditions (Tortonese, 1960;Fredj and Maurin, 1987;Bouchet and Taviani, 1992;Quignard and Tomasini, 2000).
The abrupt reduction of abundance, biomass and mean fish weight at depths greater than 1200 m and the significant decrease in species richness and diversity across the whole depth range indicate the existence of a deep-sea ichthyofauna dominated by small species living in a food-scarce environment in which some large mobile fishes are widespread. According to studies in the Atlantic ocean (e.g. Dayton and Hessler, 1972;Isaacs and Schwartzlose, 1975;Haedrich and Rowe, 1977;Mahaut et al., 1990;Merrett and Haedrich, 1997), the smaller fish species, such as tripodfish and grenadiers, are adapted to a food-limited regime, feeding on epibenthic and benthopelagic organisms, while the larger species, such as sharks, depend on food arriving from the pelagic domain, such as dead animals descending from above, midwater organisms that approach the bottom and juveniles of bathyal and abyssal species that return to deep water during their ontogenetic migration. In the deep sea the bottom can be considered as a wide barren area where distribution of food is patchy. The large and more motile animals are generalist feeders which utilise a broad spectrum of prey (Hureau et al., 1979;Haedrich et al., 1980). Haedrich and Rowe (1977) reported that the need to search a wide area for a few randomly occurring food items and the lower metabolic demand per unit weight in large animals should select for large efficient seekers in the deep-sea megafauna. Albertelli et al. (1992), casting traps at 4500 m in the eastern Mediterranean, did not find small scavenger crustaceans and echinoderms but only relatively large swimming prawns and mysids, and specimens of Chalinura mediterranea. In the opinion of these authors, the high activity and the long range foraging needed in an unpredictable food environment would exclude or drastically limit the smaller groups of crustaceans. Jones et al. (2003) recorded only the caridean shrimp Acanthephyra eximia and the macrourid C. mediterranea below 4000 m, indicating that the presence of these two species at depths at which the other species were absent might be due to a combination of their small size and mobility as well as their opportunistic and euryphagic feeding strategy (Cartes and Maynou, 1998;Carrassón and Matallanas, 2002).
Many studies on feeding habits in deep-sea fish species have indicated the presence of both active predators such as C. coelolepis and E. spinax feeding on a wide range of benthic and free swimming organisms, and species such as B. mediterraneus with reduced mobility and a microphagous filtering diet (e.g. Hureau et al., 1979;Haedrich et al., 1980;Gordon, 1984a, 1984b;Carrassón and Matallanas, 1990;Mahaut et al., 1990;Carrassón et al., 1992;Merrett and Haedrich, 1997;Carrassón and Matallanas, 2001). In addition, a degree of overlapping between these two extreme trophic strategies seems to be found in other species belonging to the Macrouridae, Moridae and Bythitidae families, which feed on food occurring within the benthic boundary and at the bottom interface (Isaacs and Schwartzlose, 1975;Geistdoerfer, 1978;Macpherson, 1979;Gordon, 1984a, 1984b;Carrassón and Matallanas, 1990;Mahaut et al., 1990;Carrassón and Matallanas, 2002). According to these studies, the present results would indicate that: 1) the differences in the ichthyofauna structure between the Balearic Sea and the Ionian Sea are more marked on the upper and middle slope than on the lower slope; and 2) in the Mediterranean lower slope assemblage, which is very little diversified, several life styles coexist: the smaller and medium sized species feeding mainly on benthopelagic plankton or strictly benthic organisms are widespread throughout the basin as typical species of food-limited assemblages; the large car-nivorous species, roaming a vast area searching for randomly occurring large food items, seem to be differently distributed between the Balearic and Ionian Sea. These findings indicate that our knowledge on the distribution of deep-sea organisms and its causes is still fragmented.