Influence of trawl type on the composition and diversity of deep benthopelagic fish and decapod assemblages off the Catalan coasts (western Mediterranean)

a series of 57 (600-750 m) and 32 (1184-1320 m) hauls were compiled to compare data from four trawl types in analyses of the species composition and diversity of deep benthopelagic fish and decapods off the Catalonian coasts (western Mediterranean). two 1-warp trawls (otsB-14 and Mts-25) and two 2-warp trawls (a commercial trawl [Ct] and a GoC73) were used, from hauls performed between 1988 and 2007 were compiled. at 600-750 m the otsB-14 hauls were separated in multi-dimensional scaling (MDs) from hauls performed with Mts-25, Ct, and GoC73. ordination of hauls along DiM1 of MDs was significantly correlated (spearman’s r) with mesh size at codend, bridle length, and vertical opening of trawls. When Ct and GoC73 were compared with Mts-25, the most dissimilar species were Aristeus antennatus, Galeus melastomus and Trachyrhynchus scabrus. a particular case was A. antennatus, with a significantly higher abundance in Mts-25 catches than the rest of the trawls. at 1200-1300 m the otsB-14 and Mts-25 hauls were segregated in MDs analysis. DiM2 was correlated with the sampling period (month), indicating a temporal effect on haul composition. at 12001300 m the main dissimilarities between otsB-14 and Mts-25 were marked by Alepocephalus rostratus and A. antennatus (caught with Mts-25), and by Cataetyx alleni, the notacanthidae Notacanthus bonapartei and Polyacanthonotus rissoanus, and Plesionika acanthonotus (caught with otsB-14). as a conclusion, otsB-14 showed a significantly higher abundance in catches than Mts-25, Ct and GoC73, and a higher S (number of species) and J (equitability) than Mts-25. on the lower slope seasonality influenced trawl comparisons; however, it is likely that Mts-25, with higher vertical height than otsB14, caught the benthopelagic fish A. rostratus more efficiently, explaining the higher biomass from Mts-25 there. Mts-25 positively selected A. antennatus, in parallel to a negative catchability for other species.

the deep Mediterranean is considered to be a natural laboratory (Margalef, 1985), due to its thermal (13ºC) and saline (38 psu) stability below depths of only 150 m in the water column, though long-term changes in oceanographic conditions have also been reported (Vargas-yáñez et al., 2009).thus, environmental variability in areas such as the Catalan sea slope seems low in comparison with other deep-sea areas. in addition to environmental variability, the type of collecting gear is another factor affecting the study of intra-and interannual changes in the composition and distribution of demersal bathyal assemblages (Merrett et al., 1991a;Gordon and Bergstad, 1992;Bertrand et al., 2002a).selectivity among trawls depends on many variables, with mesh-size, mouth area, towing speed and the number of warps towing the net being the most important.trawl comparisons have been analysed (e.g. by side-by-side trawling experiments) in shallow waters (Wilderbauer et al., 1998;szalay and Brown, 2001), basically for fisheries purposes.similar studies have also been performed at bathyal depths (Pearcy et al., 1982;Merrett et al., 1991a, b;Gordon and Bergstad, 1992), more focused on ecological implications relating to biomass and size vs. depth distributions (Haedrich et al., 1975;Pearcy et al., 1982).
ties, catching mobile species more efficiently.this trawl was mainly introduced to improve catches of the red shrimp Aristeus antennatus, the main target species of deep-sea fisheries in the area, incorporating features of conventional commercial gear used by red-shrimp fishermen off the coasts of Catalonia (sardà et al., 1997).other sources of ecological information have been obtained from the use of agassiz trawls (abelló andValladares, 1988), commercial trawlers (abelló et al., 1988), and monitoring studies performed within the framework of MEDits-spain cruises (see Bertrand et al., 2002b;Cartes et al., 2002).MEDits cruises have been performed along the spanish coasts yearly in spring (May-June) since 1994, operating to ca. 800 m depth under the coordination of the instituto Español de oceanografía (iEo) as in the case of commercial trawlers, MEDits cruises used a GoC73, a 2-warp system with the general design of a conventional commercial otter trawl (Bertrand et al., 2002a, Fanelli et al., 2007).
the aim of this study was to establish a comparison among trawls commonly used in studies on megabenthos in the deep, western Mediterranean in terms of: 1) species composition; 2) abundance/ biomass; 3) species richness/diversity.We further discuss the biases introduced by each trawl in ecological aspects, such as zonation patterns and species distributions, and in the study of megabenthic biodiversity in bathyal habitats.

The sampling area
We chose for our comparative study the area of the Catalonian coasts between Barcelona and the Ebro Delta, which has been the subject of many studies over the last 20 years.these sites are not under the direct influence of upwelling processes (local winds blowing in the Gulf of Lyons) (Fig. 1) and associated geological features (e.g.advective fluxes above large canyon: Palanques et al., 2005).to remove the effect of depth gradients on assemblages, we compared hauls performed in two fixed bathymetric strata: 1) the mid-slope between 600 and 750 m, depths commonly worked by red-shrimp trawlers (Maynou, 2008); and 2) the lower slope between 1184 and 1320 m, depths which are free of any trawling activity.the features of the trawls used are listed in table 1. together with the two 1-warp trawls used by iCM-CsiC teams (otsB-14, Mts-25), and the GoC73 used by iEo on MEDits cruises, we introduce data from a conventional trawler, the FV Stella Maris III operating in the harbour of Barcelona.the main differences among trawls were in the mesh size at the codend (6 to 40 mm), the length of bridles (8 to 150 m), and the vertical height or opening (1.2 to 3 m) (table 1).

Data analysis
Fish and decapods were counted and weighed by species and classified to species level in all cases both on board and in the laboratory.
abundance and biomass (g wet weight/ha) were calculated by swept area after raw numbers (total in- dividuals collected per haul) had been standardised to individuals per hectare.Horizontal length (m) recorded by sCanMar devices mounted on wing trawls, median towing speed (knots) and haul duration (h) were used to calculate swept area.Multidimensional scaling (MDs) was performed on matrices of data similarities.two distances were used for the middle slope, the quantitative Bray-Curtis distance after log-transformation of data (Clarke and Warwick, 1995), and the non-parametric spearman's correlation coefficient, a rank-based measure.Bray-Curtis was the only distance used for the lower slope data, because both measures showed basically the same trends.Fishes and decapods were analysed together, while non-arthropod invertebrates were excluded from analyses because trawling is not a suitable method for quantifying epibenthic and infaunal invertebrates.as a consequence, variations found in epifaunal-infaunal catches cannot be due only to natural variations, but depend on the unpredictable behaviour of trawls collecting these non-swimming, often buried, species.PErManoVa analyses (one factor, trawls; 999 permutations) were performed on a fish and decapod resemblance matrix (Bray-Curtis distance) to test the significance of groups formed in MDs analyses.siMPEr was performed to establish the contribution of each species to dissimilarities between the types of trawl.Factors (trawls) were the four types of trawl employed (Ct, otsB-14, Mts-25, GoC73).
spearman's correlations were calculated between MDs dimensions and a number of variables likely a priori to influence the species composition per trawl.the following variables were considered: i) technical variables: mesh size at codend, trawl height (vertical opening) and bridle length (see table 1) ii) natural variables: latitude (Lat) and longitude (LonG) to define the fishing grounds or areas, time of the day (h, GMt), intrannuality (month) and interannuality (year) of samples.intrannuality of surveys was distributed as follows: BatHos (June-July, october 1988), sPErMa (october 1989), ZonaP (May 1992), GEoDELta (May 1994), rEtro (april-December 1991, March-July 1992), BatMan (March 1994, June 1995) and BioMarE (January 2006 to February, april-May, June-July and october 2007).a rank order was assigned per month as a function of increasing duration of daylight at the latitude studied (e.g. 1 for December, 12 for June).MEDits were performed between 2002 and 2007, always in May-June.
iii) Measures of species diversity: number of species (n), shannon index (H') and evenness (J).
Differences in biomass per trawl were tested by 1-way anoVa (after log-transformation of data) for total biomass and for the main dominant species of fishes and decapods.Mean weight per haul (biomass/number) was calculated for key species having the greatest contribution to dissimilarities (e.g.Aristeus antennatus; Alepocephalus rostratus), and differences among trawls were tested by 1-way anoVa.
Measures of diversity (H, shannon-Weaver index), evenness (J, Pielou) and species richness (S) per haul were calculated.Differences of mean S, H and J were tested among gear types by 1-way anoVa.We performed the Wilcoxon paired- match test to establish whether the differences between each set of pairs of species abundance per trawl were significantly different from the expected 50% of switches between trawls in the absence of significant bias.We used Mann-Whitney tests to evaluate paired comparisons of species abundance from gear to gear.all the analyses were performed using PriMEr6 & PErManoVa+ (Clarke and Warwick, 1995;anderson, 2001) and statis-tiCa 7.0 software.rEsuLts

600-750 m depth range
Based both on Bray-Curtis and on spearman's r distances, MDs showed similar aggregations of hauls as a function of the trawl type (Fig. 2a).Based on Bray-Curtis, the otsB-14 hauls (on the left side of the plot) were separated from the Mts-25, Ct and GoC73 hauls.the six hauls performed with commercial trawls (Ct) and the five GoC73 hauls were grouped together.on DiM 1 (horizontal axis), the Mts-25 hauls were on average closer to the Ct hauls, and to a lesser extent to GoC73, than to the otsB-14 hauls.average trawl composition differed significantly between the 4 trawls used (PErManoVa test: pseudo F=7.44, P=0.001; post hoc tests, P<0.01 between all paired comparisons).Based on spearman's rank correlation, segregation between the otsB-14 and GoC73 hauls was not as evident as using Bray-Curtis; GoC73 hauls occupied an intermediate position between the otsB-14 and Mts-25 hauls in MDs.Ct hauls were clearly aggregated and separated from the other hauls (Fig. 2B).2). the biomass per swept area obtained with otsB-14 was slightly less than for the other three trawls, but without any significant difference among the trawls (1-way anoVa: F=0.36; P=0.78) (table 2). a 1-way anoVa performed on S (species richness), showed higher S for otsB-14 than for Mts-25 (F=4.06,P=0.01; scheffé test, P=0.04) (table 2), while differences between Mts-25 and GoC73 s were close to the significance level.shannon's H' diversity was also higher for otsB-14 than for Mts-25 (scheffé test, P=0.013; table 2), while no significant differences among trawl type were obtained for Pielou's (J) equitability (table 2), 1200-1300 m one-way anoVa performed on S, H' and J did not indicate significant differences between the Mts-25 and otsB-14 trawls.significant differences were not found for total density (D), although D in Mts-25 hauls was higher and close to statistical significance (P=0.05).Biomass was significantly higher in the Mts-25 hauls (table 2).

600-750 m
Bray-Curtis MDS.spearman's rank correlations indicated significantly positive relationships between DiM1 values from the MDs (Fig. 2) and mesh size (r=0.718;p<10 -6 ), which is consistent with the respective positions of the Ct hauls (mesh size = 40 mm) and GoC73 (mesh size = 20 mm) and the otsB-14 hauls (mesh size = 6 mm) on opposite sides of the plot. in parallel, we found similarly significant relationships (r=0.791;P<10 -6 ) regarding other technical features of trawls (bridle length, opening height), with the highest trawls having the largest bridles and otsB-14, with the lowest vertical opening and the shortest bridle, occupying opposite positions in the MDs plot (Fig. 2a).Furthermore, the trawl-characteristic variables H' (r=-0.384;p=0.003) and J (r=-0.513;P=4• 10 -5 ) were negatively correlated with DiM1; that is, diversity was higher in the otsB-14 hauls.Finally, DiM 1 was almost significantly correlated (P=0.05) with Lat. this is because GoC73 and Ct were exclusively performed off Barcelona, at the northernmost latitudes in our sampling area.the latitudinal effect was weak because these southernmost hauls (GE hauls, off the Ebro Delta) were not situated to the left along DiM1.DiM2 positions were not correlated with any variable.
no significant correlations between year and month of sampling and DiM of MDs analyses (either for Bray Curtis and spearman distances) were found.

1200-1300 m
Bray-Curtis MDS.For the lower slope trawls, we did not test mesh size, bridle length or opening height as explanatory variables, because we had only one source of variation (Mts-25 vs. otsB-14 trawls).species number, S, was positively correlated with DiM1 (Fig. 3; r=0.389;P=0.02, i.e. higher S on average for hauls situated to the right).this does not mean, however, that S was higher among Mts-25 hauls, because Mts-25 and otsB-14 hauls were clearly separated along DiM1, whereas S was not significantly correlated with this axis.DiM1 was also negatively correlated with J (r=-0.409;P=0.02).
DiM2 was correlated with month of sampling (r=-0.356;P=0.04).Most Mts-25 hauls were in March-april, while otsB-14 hauls were mainly performed in June-July and october.no significant correlations were found regarding the year of sampling or MDs dimensions.

SIMPER results: species contribution by trawl types
Considering only those species contributing ≥5% of d, higher similarities between species composition (less d) were found in the comparison of GoC73 vs otsB-14 (d=30.8%)and Ct vs Mts-25 (d=42.5%). in both cases, higher similarities were found in paired comparisons between 1-and 2-warp trawls than between the two 1-warp trawls (d=43.1%)or the two 2-warp trawls (d=59.0%).

600-750 m
When the trawls with the two largest mesh sizes at the codend (Ct and GoC73) were compared with otsB-14, the species contributing most to dissimilarities (d) according to siMPEr analysis (table 3) were small crustaceans such as Calocaris macandreae, Munida tenuimana and Sergestes arcticus.When Ct and GoC73 were compared Mts-25, however, the most dissimilar species were not small species, but large ones such as Aristeus antennatus, Galeus melastomus and Trachyrhynchus scabrus.a particular case was Aristeus antennatus, for which d was greater between Mts-25 and the large 2-warp trawls (Ct and GoC73, respectively 15.2-15.7%)than between Mts-25 and otsB-14 (8.1%).
Comparing catches of the two 1-warp trawls, a significant number of species showed greater abundance in the otsB-14 hauls (n=19) than in the Mts-25 hauls (n=4) (Wilcoxon paired-matched test was significantly different from the expected 50% in the absence of any significant effect, P>0.001).no significant trends were found between the two 2-warp systems.
the number of species collected more effectively by Mts-25 was not significantly different from the number collected better by otsB-14 (Wilcoxon paired-matched test).

DisCussion
our approach was based on trawl series performed in two relatively narrow depth ranges (600-750 m and 1200-1300 m) in a homogeneous area over a period of ca.20 yrs.though it was not a side-by-side trawling experiment, as performed at shelf depths (e.g. by szalay and Brown, 2001), similar approaches combining data from different surveys have been applied at bathyal depths (e.g. in the north atlantic: Merrett et al., 1991a, b) due to the limitations of performing trawl and side-by-side surveys at great depths.

Natural and human-induced variability
Previous on muddy, bathyal bottoms off the iberian coasts showed significant changes in megafaunal assemblages at the spatial mesoscale (latitude: Cartes et al., 2002) with an effect of the mainland/island factor (Cartes et al., 2004).trophic guilds of decapods and fish varied along the spanish Mediterranean coasts from Catalonia to the straits of Gibraltar, with a north to south increase in plankton feeders.also, around insular areas in the iberian Mediterranean coasts, low advective fluxes (e.g.river discharges) and the impoverishment of benthic biomass are characteristic (Cartes et al., 2004), enhancing differences between insular and mainland bathyal communities.as latitude was not correlated with species composition in our MDs analyses, we can consider the geographic area selected as homogeneous for our purpose.
anthropogenic factors, mainly fisheries, may also influence the composition of megafaunal assemblages in the study area.the trawler fleet has moved deeper (to ca.900 m) in search of Aristeus antennatus in recent years off the Catalan coasts, though the influence of trawling on communities is nil below 900 m depth (Cartes et al., 2004), and the lower slope at 1200-1300 m must be considered as a pristine area.Isidella elongata, a soft-coral forming a characteristic facies in the deep Mediterranean (Pérès, 1985), which is inhabited by important commercial species (the large red-shrimps Aristeus antennatus and Aristaeomorpha foliacea), is immediately removed from fishing grounds by trawling.Habitat generated by Isidella enhances diversity of fish and especially of decapods, some of them living as epibionts on coral colonies (author's unp.data).Geodelta survey (GE) hauls were the only ones performed in pristine areas at 600-750 m, while the rest were from fishing grounds of the red shrimp Aristeus antennatus.However, the massive occurrence of Isidella elongata colonies (author's unp.data) in GE hauls cannot influence our analyses performed only on fish and decapods (GE hauls not segregated from the rest performed on fishing grounds in MDs analyses).
other sources of variability in deep-sea communities are related to temporality (tyler, 1988). in the deep Mediterranean, seasonality strongly influences the dynamics of benthopelagic communities (Cartes, 1998;Cartes et al., 2008), trophic relationships and food consumption (Cartes, 1994;Madurell and Cartes, 2005), and the biological cycles of bathyal species, both macrofauna and megafauna (Cartes and sorbe, 1996;Fanelli and Cartes, 2008).though seasonality (= month) had no significant effect on trawls from the 600-750 m stratum, it significantly influenced the species composition at 1200-1300 m, so trends observed in that stratum can be biased by natural variability and fluctuations of biomass, in addition to trawl characteristics.seasonal fluctuations in abundance were evident in the study area on the lower slope (sardà et al., 1994).Compared with a previous study (sardà et al., 1997), in which the lack of any seasonal effect was assumed, we found some seasonal influence on species composition after comparing Mts-25 hauls (performed between 1991 and1995 mainly in March andapril) with otsB-14 hauls (performed in 1988-1989 mainly in June-July and october).

Trawl type variability
the main trend in the comparison of assemblage composition using different trawls was the sharp bias introduced by Mts-25 in the catches of Aristeus antennatus (i.e.>4 times higher than otsB-14 catches) both at 600-750 and at 1200-1300 m. this positive selectivity of red shrimp by Mts-25 was also found when it was compared with 2-warp trawls, i.e. commercial hauls.it is well known that A. antennatus has strong interannual fluctuations in abundance, and in some years it even disappears from the fishing grounds (Maynou, 2008).Landings at the harbours of Barcelona and Vilanova were especially high in 1991-1992, whereas they decreased to normal or average catches in 1994-1995 (Maynou, 2008).therefore, the peak of A. antennatus in 1994-95 using an Mts-25 trawl did not coincide with landing changes.the presumably low catches in 1992 with an otsB-14 even coincided with a peak in landings.thus, the differences observed were not related to fluctuations in red shrimp abundance but to the type of gear used.
on the lower slope a similar trend of higher catches with Mts-25 was found for the fish Alepocephalus rostratus, in agreement with previous analyses (sardà et al., 1997;Moranta et al., 1998).sampling with trawls of different opening height gives ecological information on the distribution of species close to the bottom (Engås and Godø, 1986).A. rostratus feeds on gelatinous plankton (medusae, pyrosomids: Carrassón and Cartes, 2002) at 1200 m depth, so its benthopelagic behaviour may favour avoidance of trawls with a low opening height like otsB-14.this trawl opens 1.2 m vertically (as sCanMar systems record) on average (cited as 2 m by sardà et al., 1997).Because Mts-25 opens to 2 m, we can conclude that most A. rostratus biomass occurs well above one metre from the bottom.such biases may change the apparent bathymetric location of peaks of biomass, especially in swimming or bathypelagic animals (Merrett et al., 1991a).Gordon and Bergstad (1992) have shown that fast-swimming species (e.g.squalid sharks and alepocephalids) were more abundantly caught by a wider trawl operated at higher trawling speeds.We think estimates of A. rostratus abundance mainly depend on the opening height of trawls used.
Codend mesh size is one of the main factors explaining differences in species composition using different trawls (Merrett et al., 1991a).Despite the small (10 mm) mesh of the Mts-25 codend, the greatest dissimilarities were observed when Mts-25 was compared with 2-warp trawl catches.these larger differences were not found among small species, as expected, but among medium and large-sized species (e.g.sharks and macrourids).in addition, medium-sized species, such as the fish Lepidion lepidion and the shrimp Acanthephyra eximia, were caught in significantly greater numbers using an otsB-14 than an Mts-25.this suggests that Mts-25 is selective for most species, probably because it was designed specifically for a commercial species, the red shrimp.in contrast to Alepocephalus rostratus, positive selectivity of Mts-25 for Aristeus antennatus in comparison with otsB-14 is unrelated to swimming capacity, because i) the red shrimp diet off the Catalan coasts is mainly infauna and suprabenthos (Cartes, 1994), which are distributed at 0-1 m above the bottom, and ii) similar yields were obtained for A. antennatus using an otsB-14 and 2-warp trawls (Ct/ GoC73) with opening heights of ca. 3 m.as Mts-25 does not catch smaller species/specimens than commercial trawls, it probably enhances biggerdeeper trends when it is used across wide bathymetric ranges.
in terms of species richness and diversity, otsB-14 behaved better than Mts-25 at 600-750 m, with higher S and H'.surprisingly, diversity of Mts-25 hauls was no greater than that of 2-warp trawls.this can also be explained by the highly positive selectivity of Mts-25 towards Aristeus antenna-tus (see Fig. 4), because the high dominance of a single species reduces H' values.Different trawls also increase artifactual variability when patterns found in different areas are compared, e.g. in studies of zonation.For example, Maynou and Cartes (2000) found differences in zonation for decapod crustaceans collected in the algerian Basin from patterns previously found in the Catalano-Balearic Basin by Cartes and sardà (1993); the former study was performed with Mts-25, the latter with otsB-14.Differences mainly consisted in a lack of renewal of assemblages at around 1200 m off the algerian Basin (Maynou and Cartes, 2000).Most species of decapod crustaceans are present in both the Balearic and algerian Basins (e.g. both mainland and insular areas: Cartes et al., 2004;Massutí and reñones, 2005).although different food availability in the two areas explains differences in species composition, the 1200-1300 m boundary coincided in the Catalan sea with significant changes in the population structure and biomass of Aristeus antennatus (sardà and Cartes, 1993), and other decapods.thus, biases introduced by differing trawl effectiveness, which are species-dependent, can affect the centres of gravity of species and the zonation patterns reported in the literature.
in conclusion, we have shown that at mid-bathyal depths off the Catalonian coasts otsB-14 is an effective trawl for community studies, showing significantly greater abundance estimates than Mts-25, Ct and GoC73, and greater diversity (both H' and S) than Mts-25.the otsB, often considered as an adequate megafaunal sampler at great depths (>1000 m) (rucabado et al., 1991), also gave acceptable results for biomass (non-significantly lower values than Mts-25 and the two 2-warp systems) at 600-750 m. on the lower slope seasonality influenced trawl comparisons; however, it seems reasonable to consider that Mts-25, with greater opening height than otsB-14, was better at catching the benthopelagic fish Alepocephalus rostratus.the greater biomass obtained by Mts-25 on the lower slope is mainly attributable to this species.Mts-25 positively selected the red shrimp Aristeus antennatus, and it seemed that this positive selection was parallel to negative selection for other species of fish and decapods, giving a more biased picture of species composition than the other samplers.otsB-14 seems to be the most suitable system for the study of some trends in deep-sea ecology, particularly regarding diversity.aCknoWLEDGEMEnts We deeply appreciate the of all participants in all the cruises that our analyses were based upon, especially the crew of rVs Garcia del Cid and Cornide de Saavedra, aboard which these cruises were performed, and the crew of FV Stella Maris, especially its master Mr. V. samper.this study was performed within the framework of the BioMarE project (ref. CtM2006-13508-Co2-02/Mar), and is partially based on data generated within this project.rEFErEnCEs fIg. 1. -Map of the study area between Barcelona and tarragona off the Catalan coasts.
fIg. 2. -MDs (multi-dimensional scaling) analysis for fish and decapods collected with four bottom trawls at 600-750 m. a, Bray-Curtis distance; B, nonparametric spearman's r.Labels indicate month and year of sample.
fIg. 4. -Dominant species of fishes and decapods as a function of gear at mid-slope depths (600-750 m).asterisks (**) above species indicate that differences in catchability were significant among trawls.
fIg.5.-Dominant species of fishes as a function of gear at the lower slope depths (1200-1300 m).asterisks (**) above species indicate that some differences in catchability were significant among trawls.

table 1 .
-Main comparative characteristics of the four trawls used in the study area.

table 3 .
-siMPEr analysis results by depth strata and comparing (average dissimilarity) trawl by trawl.Ct, commercial trawl.