Habitat use by the pearly razorfish, Xyrichtys novacula (Pisces: Labridae)*

Patterns of habitat use are documented by relating the abundance of different components of a fish population to specific environmental variables (Kramer et al., 1997). Each species is reliably found in some kinds of locality and not in others, depending on the habitat characteristics. Habitat characteristics are fundamental to the ecology and evolution of a species and the understanding of habitat use and habitat selection is a prerequisite for the preservation of endangered species and the sustainable exploitation of commercial stocks. The pearly razorfish, Xyrichtys novacula (L., 1758), is a labrid fish that inhabits clear shallow areas with sandy bottoms. It is a subtropical species, encountered in the Western Atlantic, from the North Carolina coast through the Caribbean to Brazil, and SCI. MAR., 69 (2): 223-229 SCIENTIA MARINA 2005


INTRODUCTION
Patterns of habitat use are documented by relating the abundance of different components of a fish population to specific environmental variables (Kramer et al., 1997).Each species is reliably found in some kinds of locality and not in others, depending on the habitat characteristics.Habitat character-istics are fundamental to the ecology and evolution of a species and the understanding of habitat use and habitat selection is a prerequisite for the preservation of endangered species and the sustainable exploitation of commercial stocks.
Palabras clave: ecología del comportamiento, Mediterráneo, granulometría del sedimento.*Received May 26, 2004. Accepted November 29, 2004. in the Eastern Atlantic, from the southern coast of Spain to south of Cape Lopez, including the Mediterranean and many Atlantic islands, at depths from 1 to 90 m (FishBase 2004: World Wide Web electronic database; available from the Internet URL http://www.fishbase.org).Its prey varies according to the geographical area.In the central Mediterranean it feeds mainly on bivalves, essentially Acanthocardium tuberculatum, and sea urchins, mostly Echinocardium cordatum, but also on gastropods and crustacea (Cardinale et al., 1997), while in the Virgin Islands it feeds mainly on gastropods, bivalves and also on polychaetes, crustaceans and other molluscs (Randall, 1967).The pearly razorfish is a protogynous monandric hermaphrodite (Bentivegna and Rasotto, 1987) with sexual dimorphism apparent in head shape, length of pelvic fin and coloration (Gomon and Forsyth, 1990).X. novacula is polygynous and haremic (Marconato et al., 1995, Cardinale et al., 1998).Females occupy small adjacent territories on extended flat sandy bottoms and males patrol and defend larger areas usually encompassing four to six female territories (Marconato et al., 1995).
Unlike fish of coral reefs or rocky substrates that take advantage of physical structures (crevices, holes, etc) for shelter, fish on sand flats have to seek other ways to protect themselves from predators.Some of the adaptations for avoiding predators in sandy habitats include burrow construction, armour, toxicity, mimicry, cryptic coloration, group living and sand-diving (Nemtzov, 1994).The pearly razorfish dives, head first, into the sand with the approach of danger (Lieske and Myers, 1994 and personal observation) and may remain buried for long.The pearly razorfish constantly uses the same 'dive sites', which are distinguishable visually by a slight difference in the coloration than that of the surrounding sand; furthermore one can poke a finger into a dive site more easily than into a nondive site area just a few cm away (Clark, 1983 and personal observation).This burying defence strategy seems to be common among most species of the genus and it has been reported for many other wrasses as well, e.g. for X. martinicensis (Victor, 1987), X. pentadactylus (Clark, 1983), X. niger (Clark, 1983), X. splendens (Nemtzov, 1994).The unusual head shape of razorfish is an adaptation that enables sand-diving.Sand-diving apparently presupposes that the soft substrate is adequate for easy 'diving' at a sufficient depth that assures protection from predators.Furthermore, no impediments in respiration should be caused by the sand when the razorfish remains buried for long.Thus, it seems that the distribution and abundance of the pearly razorfish on soft substrates may be restricted by the characteristics of the sediment.In this study the population density of the pearly razorfish was estimated in eastern Mediterranean areas and related to specific environmental variables, laying emphasis on the sediment granulometry.

MATERIALS AND METHODS
The population density of the pearly razorfish was estimated by visual census with SCUBA diving in Greek coastal waters (Eastern Mediterranean) and at depths from 3 to 25 m.The survey was conducted in randomly chosen areas with soft sediment.The X. novacula individuals were counted within strip transects, ranging between 1600 and 2080 m 2 .The transects were outlined in a way similar to that described in Katsanevakis and Verriopoulos (2004).A total of 89 transects were outlined and measured in different sites in Eastern Mediterranean (Fig. 1).The survey lasted one calendar year, from July 2001 to June 2002.
A large percentage of pearly razorfishes were found hovering over their dive site.Those pearly razorfishes that were far from a dive site were usually made, by the presence of the diver, to move to the nearest dive site and hover over it.Razorfishes are territorial, do not move farther than a few meters from a dive site (Clark, 1983) and although the presence of the diver may force them to dive into the sand, this does not happen unless the diver approaches too close (less than ≈ 1.0-1.5 m).Thus, in most cases the diver had the time to spot and record the fish before it dived into the sand.Consequently, underestimation of X. novavula density due to failure to record all individuals is considered to be small.
The depth range of each transect (maximumminimum depth) was recorded with a dive computer, with an accuracy of 0.1 m, and in every case was less than 5 m.The mean value of the minimum and maximum depth of each transect was considered as the transect depth (D).
A 250 ml sample of the surface sediment (upper 5 cm) was taken from each transect.Particle size analysis of the sediment samples was conducted according to Buchanan (1984) and for each sample the median diameter Mdφ and the quartile deviation QDφ were calculated as measures of the central ten-dency and the degree of scatter of the granule size frequencies respectively.
The presence or absence of Acanthocardia tuberculata and irregular sea urchins in the surveyed transects was recorded.As these species are endobenthic, accurate estimation of their abundance is not feasible, thus only presence/absence data, based on empty shells was recorded.
General Linear Model (GLM) methods were used (Glantz and Slinker 2001) to identify the asso-ciation between X. novacula density and the observed environmental or biotic variables, giving emphasis in sediment granulometry.Initially, a General Linear Model of the following form was calculated, with least squares method and assuming a normal error distribution: , ( 1 Densities were fourth-root transformed in order to stabilize variance and produce fairly straight lines on the normal probability plots (Glantz and Slinker 2001); untransformed densities produced curves on the normal probability plot with one inflection, indicating that the distribution of the residuals was skewed.Variables S 1 , S 2 , S 3 are dummy variables, used to encode the effect of season, following the 'effects coding' approach, according to Glantz and Slinker (2001).Specifically, when the season is {autumn, summer, spring, winter} respectively then S 1 ={1,0,0,-1}, S 2 ={0,1,0,-1} and S 3 ={0,0,1,-1}.P A and P E are dummy variables (with 'effects coding') representing the presence/absence of Acanthocardia tuberculata and irregular sea urchins respectively.When the data are unbalanced, as in our case, 'effects coding' is essential for obtaining the correct sums of squares in the model (Glantz and Slinker 2001).The 'StatGraphics Plus v.4.0' (Statistical Graphics Corp.) software was used for the analysis.Marginal Sums of Squares (Type III) were used to test the significance of each regression coefficient.A residual analysis was conducted, according to Glantz and Slinker (2001), to check whether the results were consistent with the model assumptions.

RESULTS
The density of Xyrichtys novacula ranged from 0 to 20.0 individuals per 1000 m 2 .Mdφ varied from -1.0 to 4.6 and QDφ from 0.2 to 1.63.The raw data of all measurements are given in Table 1.
After the fourth-root transformation of densities, the residuals plotted against any independent variable or against the observed dependent variables showed no deviation from the constant variance assumption and the normal probability plots of the residuals were reasonably linear, indicating no substantial deviation from normality.The residuals showed no trend, curve or other systematic variation and in the scatterplots of the transformed densities against any of the independent variables there was no indication of nonlinearity; thus the linearity assumption, inherent in the GLM model, may be considered valid.There was no studentized-residual greater than 3.0 (maximum studentized-residual was 2.72).In every case, there was no leverage value greater than 3 times the average leverage and no data points with unusually large values of Cook's distance (no Cook's distance greater than 0.0085); thus there were no outliers or influential points.The variance inflation factor of all the regression coefficients was less than 2.4 in every case, thus there is no significant multicolinearity among the variables (Glantz and Slinker 2001).
The results of fitting the GLM described by Equation 1 are presented in Table 2.The induced model was: , where density is given in individuals per 1000 m 2 .
Among the factors of the model, only Mdφ and QDφ were significant (Table 2).If only these two factors are kept in the model, it becomes (after recalculation of the new coefficients with least squares): The adjusted R 2 of the reduced model is 43.9% and the standard error of the estimate is 0.556, which are both quite close to the respective values of the full model (Table 2).
The coefficient of Mdφ is significantly negative (p<0.001), indicating that the finer the sediments, the less abundant is X. novacula.The coefficient of QDφ is also significantly negative (p=0.008), indicating that the better sorted is the sediment the higher is the X. novacula density.This is enhanced by the Student-Newman-Keuls multiple comparisons tests (Table 3); it is obvious that the highest X. novacula densities occur in moderately or well sorted, coarse or very coarse sand.

Density
Md QD

DISCUSSION
The sediment characteristics represent a major factor in regard to the distributional patterns of the bottom fauna (Bacescu, 1972).The sediment characteristics, specifically the median diameter Mdφ and the quartile deviation QDφ, seem to be significant determinants for the distribution and abundance of Xyrichtys novacula.It might be argued that the significant association between the X. novacula density and the sediment characteristics does not imply a causal link, as this is an observational study and not a manipulative experiment (Glantz and Slinker, 2001).The abundance and composition of prey may depend on sediment characteristics, thus the X. novacula density dependence on sediment granulometry might be partly indirect and caused by a direct dependence on prey abundance.However, no solid conclusion may be made on whether the dependence of X. novacula density on sediment granulometry was direct (due to the sand-diving behaviour of the species ) or partly indirect (due to the dependence of prey abundance on sediment characteristics).Nevertheless, the sand-diving behaviour of X. novacula necessitates the existence of appropriate substrate and indicates a strong link between X. novacula abundance and sediment characteristics.
We found that X. novacula density is not related to the presence or absence of Acanthocardia tuberculatum or of irregular sea-urchins, which made up 90% of the volume of prey of X. novacula in the Tyrrhenian Sea (Cardinale et al., 1997).Many areas with no sign of A. tuberculata or irregular seaurchins (measurements 10, 11, 12, 13, 50, 57, 64, 80, 82 in Table 1) were abundant with X. novacula and some areas with abundance of both A. tuberculata and irregular sea-urchins had zero X. novacula density (measurements 17, 27, 30, 63 in Table 1).Comparing the stomach content analyses of Randall (1967) and Cardinale et al. (1997), it is concluded that the prey of X. novacula may vary from site to site.Furthermore, Cardinale et al. (1997) found similar diversity of the ingested food of X. novacula, as compared to that of the environment, revealing the euryphagic nature of the species.Thus, it seems that the abundance of specific prey species does not affect the abundance of X. novacula, although the total prey abundance probably does so.
X. novacula density was not significantly correlated to depth, in the depth range of this study.However, sandy bottoms, as a rule, are limited to depths down to the continental shelf in contrast to sediments with silt or clay that are found at all depths (Bacescu, 1972).Thus, the depth range of X. novacula (1-90 m) probably occurs mostly because of the corresponding depth range of sandy bottoms.Cardinale et al. (1998) stated that no X.novacula was caught in winter either by their sampling gear (fishing lines) or by using traditional fishing methods and they support the theory that during the cold season the species spends most of the time buried in the sand.The present study, though, presents evidence against this hypothesis.No significant seasonal variation in X. novacula density was found and in many winter or early spring measurements many X.novacula individuals were counted.Specifically, from December 2001 till March 2002, X. novacula was found in 10 of the 22 areas measured (measurements 50, 57, 60, 64, 65, 67, 68, 69, 70, 71 in Table 1) with a highest density of 7.7 individuals per 1000 m 2 in measurement 50.The mean sea temperature during these 10 measurements varied from 11.9°C to 16.0°C.Although a reduced activity during winter might be possible, there is no indication that during the cold season X. novacula spends most of the time buried in the sand.The fact that pearly razorfishes do not get caught by fishing gear during winter might be due to a behavioural change in winter, or, in other words, the summer reproductionrelated behaviour might render the species more vulnerable to fishing.

)
FIG. 1. -Map with the 89 transects of the study.

TABLE 1 .
-The raw data of the 89 density measurements of Xyrichtys novacula in this study, together with the values of the 6 predictive factors of Equation 1. d, density (ind/1000 m 2 ); P A and P E are presence (1) -absence (-1) data; Sp, spring; S, summer; A, autumn; W, winter.

TABLE 2 .
-Summary of the results of fitting a general linear model relating the pearly razorfish densities to the 6 predictive factors of Equation 1. Mdφ: Median grain diameter, QDφ: Quartile Deviation of grain distribution, D: Depth, P A , P E : presence/absence of Acanthocardia tuberculata and irregular sea urchins respectively (* p<0.05, ** p<0.01, *** p<0.001).

TABLE 3 .
-The results of Student-Newman-Keuls multiple comparisons for Density 1/4 of Xyrichtys novacula among the 6 Mdφ classes and the 3 QDφ classes, together with the mean value of each class.