Population structure of the pearly razorfish , Xyrichtys novacula ( Actinopterygii : Labridae ) , in sand-seagrass mosaics : spatial variation according to habitat features and sampling techniques

1 Grupo de investigación en Biodiversidad y Conservación, Facultad de Ciencias del Mar, Universidad de Las Palmas de Gran Canaria, Campus de Tafira, 35017 Las Palmas de Gran Canaria, Canary Islands, Spain. E-mail: fesprod@gobiernodecanarias.org (FE), rharoun@dbio.ulpgc.es (RH), ftuya@yahoo.es (FT) 2 Grupo de investigación en Ecología Marina Aplicada y Pesquerías, Facultad de Ciencias del Mar, Universidad de Las Palmas de Gran Canaria, Campus de Tafira, 35017 Las Palmas de Gran Canaria, Canary Islands, Spain. E-mail: emap.raul@gmail.com (RTP), pepe.solea@ulpgc.es (JAG)


INTRODUCTION
Habitat structure routinely affects the distribution and structure of nearshore fish assemblages (Gratwicke andSpeight 2005, Tuya et al. 2011).A large body of literature has covered this research topic.Studies connecting patterns in the abundance of soft-bottom fishes with the structure of the habitat are, however, fewer in number.Soft bottoms are colonized by seagrass meadows, which largely increase the heterogeneity and complexity of the habitat relative to nearby unvegetated bottoms (Boström et al. 2006, Gullström et al. 2008, Hensgen et al. 2014).A large number of fish species inhabit both types of habitat, which are typically juxtaposed in mosaics within landscapes (Ferrell and Bell 1991, Gray et al. 1996, Horinouchi 2009).Different patterns in habitat use by nearshore fishes have been described there, including fish species that are more abundant in seagrass meadows, fish species more abundant on sand bottoms, and species with similar abundance in both types of habitat, even varying at different spatial and temporal scales (Ferrell and Bell 1991, Tuya et al. 2005, Horinouchi 2009).Providing insight into patterns of fish habitat use is particularly relevant for fish that are commercially exploited, which can be described in terms of varying patterns in abundance, biomass, size structure and sex ratio that are intrinsically linked with the specific peculiarities of the habitat.
The pearly razorfish, Xyrichtys novacula (Linnaeus, 1758) (Actinopterygii: Labridae), is a benthic, protogynous hermaphrodite (Bentivegna and Rasotto 1987) distributed in warm latitudes of the Atlantic Ocean, from North Carolina (USA) to Brazil, and from the southern Iberian Peninsula to Gabon, including the archipelagos of the Azores, Madeira, Canaries, Cape Verde and São Tomé as well as the Mediterranean Sea (Froese and Pauly 2015).In the Mediterranean region, this fish is a highly prized species, particularly in the Balearic Islands (Box et al. 2009, Alós et al. 2012), where it reaches top market prices (Beltrano et al. 2006).In the Canary Islands, however, captures of this fish are low, because it has never been a target for the local cuisine.In the Canarian small-scale fisheries, it is captured incidentally through hooks, fish-traps and trammel nets that target other fish species (Franquet and Brito 1995).Along its distributional range, several aspects of the biology and ecology of X. novacula have been previously studied, including reproduction (Bentivegna and Rasotto 1987, Marconato et al. 1995, Cardinale et al. 1998, Candi et al. 2004), juvenile morphology (Mercader 1991), age and growth patterns (Cardinale et al. 1998, Battaglia et al. 2010), trophic ecology (Cardinale et al. 1997, Castriota et al. 2005a,b, 2010, Beltrano et al. 2006), and habitat use (Katsanevakis 2005, Alós et al. 2012).This fish species often inhabits shallow waters (0-90 m depth, Fischer et al. 1987, Franquet and Brito 1995, Froese and Pauly 2015) on sandy bottoms (Mercader 1991, Brito et al. 2002, Katsanevakis 2005, Tuya et al. 2005, Alós et al. 2012), frequently adjacent to seagrass meadows and coral reefs (Schneider 1990).In turn, it may thrive in seagrass meadows (Fischer et al. 1987, Mercader 1991, Riera and Linde 2001, Castriota et al. 2005b, Box et al. 2009), particularly those dominated by the genera Cymodocea and Zostera (Brito et al. 2002, Castriota et al. 2005a, Tuya et al. 2005, Espino et al. 2011a,b), and less frequently in Posidonia beds (Beltrano et al. 2006, Box et al. 2010).However, there are no specific studies describing patterns of variation in the habitat use by this species in seagrass meadows, although some studies on seagrass fish assemblages, mainly in Cymodocea and Zostera, have been performed (Guidetti 2000, Guidetti and Bussotti 2002, Guidetti et al. 2002).
The goals of this study were two-fold.Firstly, we aimed to describe the spatial and temporal patterns of variation in the abundance, biomass and population structure (size structure and sex ratio) of this species inhabiting sand-seagrass mosaics through two complementary sampling techniques.Secondly, we sought to analyse whether biotic (seagrass shoot density, leaf length and meadow cover) and abiotic (sediment composition and particle size) structural elements of the habitat helped to explain variation in patterns of abundance and biomass of this species.

Study area
Three locations dominated by seagrass meadows of Cymodocea nodosa (Ucria) Aschers.were selected at Gran Canaria Island (Fig. 1); two locations located on the southeast coast, ca. 2 km apart, and one located to the southwest of the island, ca.50 km apart.There is no significant variation in oceanographic conditions (e.g.seawater temperature, salinity and chlorophyll a concentrations) at the coast between the southeastern and southwestern side of the island (Tuya et al. 2014a), and the pool of fish species is the same on both sides of the island (Espino et al. 2011a).The oceanographic conditions are characterized by the northeastern trade winds and the Canary current, which flows towards the southwest.Sea surface temperature typically ranges from 18ºC in winter to 24ºC in summer (Navarro-Pérez and Barton 2001).
The area covered by these seagrass meadows ranges between 98417 and 261550 m 2 at 10-18 m depth (Table 1, Tuya et al. 2014a).Each location was visited seasonally four times through an entire annual cycle: in February 2011, May 2011, August 2011 and November 2011.The dates were separated to encompass conditions encountered throughout an entire year.At each location, fish were sampled at two sites selected randomly, hundreds of metres apart.

Fish sampling
Pearly razorfish were sampled using two complementary sampling techniques: underwater visual censuses (hereafter UVCs) and seine nets (hereafter SNs).At each site, 12 replicated 25-m-long and 4-m-wide transects were firstly laid out randomly during daylight hours, 10:00-14:00 h.Transects were carried out by the same two SCUBA divers (6 each diver) simultaneously, surveying different areas to avoid fish being counted twice.The abundance (total number of individuals) and size (to the nearest centimetre of total length) of all individuals of Xyrichtys novacula were recorded on waterproof forms.In the study area, this procedure provides optimal precision and accuracy to account for the abundance and size structure of both rocky-reef and seagrass fishes (Tuya et al. 2004, 2005, 2006, Espino et al. 2015).Biomasses of fish counted through UVCs were calculated using an unpublished length-weight relationship (W=0.0054L 3.3602 , R 2 =0.8576, n=124).
Secondly, six replicated 25-m-long and 4-m-wide trawls were carried out using a seine net on approximately the same transect lines where the UVCs had been previously carried out.Distance between trawls varied between 20 and 30 m.Hence, each trawl was considered as an independent replicate.This technique has been successfully applied in the study area, capturing small-sized fishes (Espino et al. 2011a(Espino et al. ,b, 2015)).All fish captured by the SNs were preserved in a 10% formalin/seawater solution and carried to the laboratory, where the total length (cm, TL±0.1 cm) and weight (g, W±0.001 g) were recorded.For both sampling techniques, male and female fish were discriminated by the morphological criterion according to Oliver and Massutí (1952), Bentivegna and Rasotto (1987) and Mercader (1991).

Structural elements of the habitat
To test for predictive relationships between the physical structure of the habitat and the abundances and biomasses of Xyrichtys novacula, three biotic descriptors at each transect line were recorded: (1) shoot density (by counting seagrass shoots in six 25×25 cm quadrats that were deployed in the middle of each transect line), (2) leaf length (by measuring the average leaf length in 20 shoots selected randomly within each quadrat), and (3) seagrass cover during UVC performance (by registering the distance under a 1 cm × 25 m flexible line transect covered by the seagrass Cymodocea nodosa to the nearest centimetre and subsequent calculation of % cover, n=12) (Barberá et al. 2005, Espino et al. 2011a, 2015) (Table 1).The sediment characteristics, i.e. percentages of gravel, sand, silt, and mean diameter of particles (D 50 ), for each site within each location (n=3) were obtained from the Ecocartographic Study of the South Coast of Gran Canaria Island (Ministerio de Medio Ambiente 2002) (Table 1).

Statistical analysis
The abundance and biomass of the razorfish were partitioned by means of three-way ANOVAs that tested for differences between times (seasons), locations, and sites within locations.The model incorporated the Table 1.-Physical structure of the three locations dominated by Cymodocea nodosa seagrass meadows.Surface (m 2 ), depth ranges (m) and meadow type are provided for each location.Density (number of shoots m -2 ; mean±standard error), leaf length (cm; mean±se), cover (%; mean±se), gravel (%), sand (%), silt (%), and mean diameter of particles (D 50 , mm) are also provided for each site within each location.(Sachs 1982, Sokal andRohlf 2012).A linear regression model was fitted to pairwise abundances and biomasses obtained by each sampling protocol at each site and season (n=24) to test for a predictive relationship between them.Multiple linear regression, using the DistLM routine via 999 permutations of the data (Anderson 2001), tested the significance of the relationships between the set of predictor variables, i.e. density of shoots, leaf length, seagrass cover, percentages of gravel, sand and silt, and mean diameter of particles (D 50 ), and the total abundance and biomass of the razorfish (untransformed data).The 'Forward' selection procedure and the AIC selection criterion were applied to select the models with the largest parsimony.
The multiple regression model explained ca.30% of the variability of fish abundance.The percentage of silt, seagrass shoot density and leaf length were the predictor variables that most contributed to explaining variability in fish abundance (Table 3A, Fig. 4A-C).For fish biomass, the model accounted for ca.32% of the total variability.The variables that most contributed to explaining variability in fish biomass were seagrass cover, seagrass shoot density and percentage of sand (Table 3B, Fig. 4D-F).

Source of variation
The multiple regression models explained ca.16% and 26% of the total variability in fish abundance and biomass, respectively (Table 5).The percentage of silt, sand, and seagrass cover were the predictor variables that most contributed to explaining variation in abun-     dance (Table 5A, Fig. 6A-B).Seagrass cover, percentage of sand and the D 50 were the variables that most contributed to explaining variability in fish biomass (Table 5B, Fig. 6C-D).

Comparison between sampling techniques
The sex ratio derived from both sampling methods differed significantly (χ 2 =17.94, p<0.001); females were more abundant in SNs than in UVCs.Though UVCs recorded a larger abundance of individuals, there was a significant predictive relationship between abundances and biomasses detected through the two sampling methods (Fig. 7A-B).

DISCUSSION
In the study region, Xyrichtys novacula is found on unvegetated sandy bottoms and seagrass meadows of C. nodosa (Brito et al. 2002, Tuya et al. 2005, Espino et al. 2011a,b).The lack of temporality for the abundance and biomass patterns of this species suggests that individuals remain stable through seasons and that this species has a strong affinity for sandseagrass mosaics (Tuya et al. 2005, Box et al. 2009).In a similar way, Katsanevakis (2005) pointed out a lack of seasonal effects on densities of X. novacula from the eastern Mediterranean Sea.Individuals of this species typically display a reduced 'home range' (<0.5 km 2 ), with a small habitat use area, i.e. a sedentary life style (Alós et al. 2012).This has been explained, at least during the reproductive period, by the complex social structure of X. novacula, a polygynous species with harem behaviour, where females occupy small territories and males patrol and defend (from other males) larger areas usually encompassing four to six female territories (Marconato et al. 1995).This territorial behaviour somehow explains the temporal stability in the patterns of abundance and biomass of this species in sand-seagrass mosaics.The results of this study do not support the hypothesis suggested in the Mediterranean Sea that, during the cold season, the species spends most of the time buried in the sand (Oliver andMassutí 1952, Cardinale et al. 1998) or migrates to deep waters (Oliver andMassutí 1952, Candi et al. 2004).
In contrast, the abundance and biomass of X. novacula varied at the spatial scales of locations and sites within locations.This means that variation in razorfish abundance and biomass was mostly determined by variation in seagrass structural attributes operating at the scale of locations and sites within locations.This species has a specific relation with the substrate, as it buries in the sediment under any sign of alarm (i.e. a predator).Moreover, this species bury into burrows during the night-time, therefore displaying a clear daily rhythm of activity (Lieske andMyers 1994, Alós et al. 2012).Hence, differences in habitat structure between locations may help to explain differences in fish abundance and biomass at this scale.Our data suggest that this fish was most abundant in continuous meadows (>90% seagrass cover) with intermediate densities of 500 to 1000 shoots m -2 (see Table 1).This was particularly the case at the location 'Cabrón', where abundances reached up to 34 ind./100 m 2 .Here, fish may bury into burrows between the complex matrix of seagrass rhizomes (F.Espino and F. Tuya, pers. obs.), because seagrass shoot density does not exceed large values, i.e. 1000 shoots m -2 , which may impede the burial behaviour of this species, e.g.large above-ground stocks usually occur with high below-ground biomass, which might reduce the penetrability of infauna into the sediment (Peterson 1982, Goshima and Peterson 2012, González-Ortiz et al. 2014).Large-sized seagrass patches with shoot densities >1000 shoots m -2 surrounded by sand corridors, e.g. the location 'Risco', may also provide a suitable habitat (Espino et al. 2011b).A large seagrass shoot density may limit the capacity of fish to dig into the  sediment.However, individuals may rapidly disperse towards adjacent sandy corridors to bury and so avoid predation.The sibling species Xyrichtys splendens Castelnau, 1855, distributed in the Caribbean Sea, also inhabits shallow sandy bottoms and seagrass meadows.Though seagrass canopies provide protection against predators, the capacity to bury into the substrate is here limited in comparison with unvegetated bottoms.
In fact, the green razorfish modifies its sand-diving behaviour in seagrass habitats (Nemtzov 1994).As a result, a trade-off between protection provided by seagrass canopies and protection derived from burial behaviour may explain the spatial variability of this fish species.
The regression models showed that the average seagrass leaf length affected the distribution of Xyrichtys novacula; a larger leaf length may increase the capacity of protection by the habitat, as has been reported for many fishes in seagrass meadows (Gullström et al. 2008, Hori et al. 2009).The type of sediment (i.e. the dominant particle size) was another important driver of the abundance and biomass patterns of X. novacula in the study region.In the Mediterranean Sea, the distribution of this species seems to rely on specific sediment features; the largest abundances were found in sediments dominated by coarse and very coarse sands (Katsanevakis 2005).Though the location 'Veneguera' has a low shoot density (<500 shoots m -2 ) and seagrass cover (~63%), fish abundance and biomass were low.This may be explained by the specific features of the sediment, including a larger amount of fine sediments in comparison with the other locations (8.125 vs 1.25-2.03%):i.e. the D 50 was lower here than at the other two locations (0.14 vs 0.42-0.39mm).This would complicate the capacity of individuals to construct burrows and to bury and dive into the substrate.This result fits with findings from the Mediterranean Sea, where the finer the sediment the less abundant is X. novacula (Katsanevakis 2005); specifically, particle grain sizes between 0.125 and 0.250 mm are not suitable for this species (Alós et al. 2012).
Another element, not considered here, that may shed light to unravel patterns in abundance and biomass of X. novacula is the availability of feeding resources.This species is zoo-benthivorous, mainly consuming crustaceans, molluscs and echinoderms that inhabit bottoms dominated by well-sorted sands and coarse sands (Cardinale et al. 1997, Castriota et al. 2005a, Beltrano et al. 2006).Soft bottoms colonized by C. nodosa offer a rich fauna of endo-, epi-and suprabenthic organisms (Tuya et al. 2001, 2014b, Herrera et al. 2014), which are a potential feeding resource for juvenile and adult fish (Castriota et al. 2005b).In the Mediterranean Sea, trophic studies have revealed that the pearly razorfish is a euryphagic species that feeds on trophic resources available in each environment (Cardinale et al. 1997) and the total prey abundance may affect the abundance of razorfish but not the abundance of specific prey species (Katsanevakis 2005).As a result, the presence of these meadows may provide an ecological advantage for X. novacula, since individuals may consume prey inhabiting sandy bottoms, as well as those directly associated with seagrass meadows (Castriota et al. 2005b).
Seagrass meadows of Cymodocea nodosa may be a suitable habitat not only for adult fish but also for juveniles, as our data have demonstrated (18.28% and 33.06% of juvenile razorfish recorded by UVCs and SNs, respectively).The range of sizes obtained by each type of sampling technique influenced the sex ratio.Still, females dominated numerically in both cases, particularly in the data obtained through the SNs.This can be explained, firstly, by a higher selectivity for small sizes (0.5-10.0 cm TL) of SNs, i.e. a larger capturability for small-sized individuals (females in our case study); this has also been observed in previous studies in the study area (Espino et al. 2011a,b).In contrast, large-sized individuals (typically males, TL>17 cm) tend to escape from the sampling gear.For UVCs, however, small-sized individuals (TL<3.0cm) are difficult to detect, while large-sized individuals (TL>17 cm) are often easier to spot (Franco et al. 2012, Tuya et al. 2014b, Espino et al. 2015).In our study, UVCs recorded a larger abundance of individuals; for mature populations of this species, the sex ratio rendered by UVCs seems to provide a good approximation to the structure of the population.Our study detected a significant correlation between data supplied by the UVCs and the trawls.However, it is worth noting that the significance of this relationship was mostly driven by large abundances and biomasses at three sites.As a result, this finding should be taken with caution.
In summary, the pearly razorfish, Xyrichtys novacula, shows spatial variation in abundances and biomasses that seem to be connected with variation in habitat structural elements, with effects independent of temporality.A balance between protection provided by seagrass canopies and protection provided by its burial behaviour seems to clarify the spatial variability of this fish species.

Fig. 1 .
Fig. 1. -Map of the Canary Islands (northeastern Atlantic Ocean) showing the position of Gran Canaria Island and the three studied locations.

Fig. 2 .
Fig. 2. -Mean abundance (A) and mean biomass (B) of the pearly razorfish, Xyrichtys novacula, at each location and time, recorded through the UVCs.Error bars indicate standard error of the means.

Fig. 3 .
Fig. 3. -Total number of individuals of the pearly razorfish, Xyrichtys novacula, for each size class and sex recorded through UVCs (A) and SNs (B).

Fig. 4 .
Fig. 4. -Relationships between total abundance (A-C) and biomass (D-F) of the pearly razorfish, Xyrichtys novacula, recorded through UVCs, and the structural elements of the habitat.

Fig. 5 .
Fig. 5. -Mean abundance (A) and mean biomass (B) of the pearly razorfish, Xyrichtys novacula, at each location and time, recorded through the SNs.Error bars indicate standard error of the means.

Table 3 .
-Results of the multiple regressions tests (sequential tests) testing for significant relationships between predictor variables and the (A) abundance and (B) biomass of the pearly razorfish, Xyrichtys novacula, recorded through the UVCs.

Table 4 .
-Results of the analysis of variance (ANOVA) testing the effects of 'Time' (a fixed factor), 'Location' (a random factor, orthogonal to the previous factor), and 'Site' (a random factor nested within 'Time' and 'Location') on the abundance and biomass of the pearly razorfish, Xyrichtys novacula, recorded through SNs.

Table 5 .
-Results of the univariate multiple regressions tests (sequential test) testing for significant relationships between predictor variables and the total (A) abundance and (B) biomass of the pearly razorfish, Xyrichtys novacula, recorded through the SNs.