Relationships between otolith size and fish size in some mesopelagic and bathypelagic species from the Mediterranean Sea (Strait of Messina, Italy)

The length-weight relationships and the regressions between otolith size (length and width) and fish length of some mesopelagic and bathypelagic fishes living in the central Mediterranean Sea were provided. Images and morphological description of otoliths (sagittae) from 16 species belonging to the families of Gonostomatidae (1), Microstomatidae (2), Myctophidae (8), Phosichthyidae (2), Sternoptychidae (2) and Stomiidae (1) were given. The length-weight relationship showed an isometric growth in 13 species. No differences between right and left otolith sizes were detected by t-test, so a single linear regression was plotted against standard length (SL) for otolith length (OL) and otolith width (OW). Data fitted well to the regression model for both OL and OW to SL, for each species (R2>0.8). These relationships offer a helpful tool in feeding studies and also provide support to palaeontologists in their research on fish fossils.


INTRODUCTION
Mesopelagic and bathypelagic fishes are species usually living in mid-water masses (Salvanes and Kristoffersen, 2001), having a large vertical distribution (Gjøsaeter and Kawaguchi, 1980) and playing an important ecological role in the energy transfer from epipelagic waters to deep environments. In fact, they carry out large diel vertical migrations, moving to-wards the upper layers at night to feed on plankton or micronekton and coming back to deeper waters during the daytime to avoid predation (Marshall, 1960).
Mid-water fishes are a key component of the pelagic nekton and form a large fraction of the deep-scattering layer (DSL) (Benoit-Bird and Au, 2003). The high biomass of mesopelagic and bathypelagic fish communities in all oceans, especially in subtropical and tropical seas (Gjøsaeter and Kawaguchi, 1980;Mann, 1984;Lam and Pauly, 2005), is an important food resource in the marine trophic web.
However, during feeding studies, the identification and quantification of these preys is frequently a difficult task: in most cases specimens are already partially or totally digested and the hard remains in stomachs, intestines, faeces and scats are the only diagnostic features that can be considered. In particular, otoliths are quite resistant to the digestion and they are an important tool for prey classification in several dietary studies (Granadeiro and Silva, 2000;. Furthermore, the diet analysis of marine mammals and sea birds requires non-invasive methods, so the examination of sagittae from faeces and regurgitated digestive pellets is often the only way to recognize the preys Duffy and Laurenson, 1983;Johnstone et al., 1990). Their importance is also documented in stomach analysis of cephalopods that use their beaks to chop preys (e.g. Watanabe et al., 2004), making them identifiable only by otolith determination.
For these reasons and thanks to their high interspecific variability, several keys and identification guides on fish otoliths have been published (Smale et al., 1995;Campana, 2004;Lombarte et al., 2006;Tuset et al., 2008). To better understand the real role of fish preys' energy contribution to a predator diet, it is important to estimate their biomass and their numerical abundance. The rebuilding of body size and prey biomass from otolith measurement is possible by applying a back-calculation: a relationship between fish length and otolith size has been widely demonstrated in many fishes (Wyllie Echeverria, 1987;Gamboa, 1991;Granadeiro and Silva, 2000;Harvey et al. 2000;Waessle et al., 2003).
Moreover, the resistance of otoliths to deterioration, due to their particular calcareous structure (a concretion of calcium carbonate and other salts deposited in a protein matrix), made it possible to use them in paleontological studies. Fossil otoliths are found in a wide spectrum of sedimentary environments and are common in many marine sediments (Nolf, 1995). By comparing them to recent reference collections, many authors have made an important contribution to knowledge of the taxonomic status of the ancient fauna of the planet and to the validity of otolith-based fossil fish species (e.g. Nolf, 1985Nolf, , 1995Girone et al., 2006).
Though their reliability and scale are proportional to the availability of information on description of current otolith species, otolith data are widely used in the fields of paleoecology, paleobathymetry, paleoclimatology, paleobiogeography and biostratigraphy (Nolf, 1995).
The aim of this paper is to provide data about the relationship between otolith size (length and width) and fish length as well as between fish weight and fish length of some mesopelagic and bathypelagic fishes living in the central Mediterranean Sea. As little information is available on these fishes, in spite of their role in the food web, this paper fills an important gap. The data offer a helpful tool for feeding studies and also provide support to palaeontologists in their research on fossils. We also supply otolith images from these fishes, in order to improve and facilitate their identification. According to the terminology used by Tuset et al. (2008), we present a brief description of the sagittae for the species that were not reported in this last atlas, which offered the first major description of otoliths of recent Mediterranean fish species. This area is known for its strong upwelling currents caused by different tidal phases of the two abovementioned basins and their intensity is regulated by lunar phases (Mazzarelli, 1909;Vercelli, 1925; De Domenico, 1987). The strong hydrodynamism and the particular meteorological conditions (strong wind from the SE) are the main causes of mesopelagic and deep fauna stranding in this area (Mazzarelli, 1909;Genovese et al., 1971;Berdar et al., 1977;Spalletta et al., 1995). The biological material was collected along the Sicilian coast of the Strait of Messina before the sunrise, in order to avoid the competition of seabirds, ants and wasps and the sun's drying effect.

MATERIALS AND METHODS
Specimens were identified following Whitehead et al. (1984Whitehead et al. ( -1986, photographed with a camera, measured by calliper to the nearest 0.1 mm (following Tortonese, 1970) and weighed to the nearest 0.01 g. In most cases the caudal fin was damaged, so the standard length (SL) in place of the total length (TL) was considered. Sagittal otoliths were removed, cleaned and stored dry with a code number.
Lengths of sagittae (OL) were determined by an ocular micrometer mounted on a Zeiss Stemi 2000-C stereomicroscope and were recorded as the greatest distance measured from the anterior tip to the posterior edge, parallel to the sulcus (Harvey et al., 2000). The width of every sagitta (OW) was determined by considering the greatest distance from the dorsal otolith edge to the ventral one, perpendicular to the sulcus.
Length-weight relationship for every species was also described using the potential function: where W is the total weight, SL the standard length, a the intercept of the regression line and b the regression coefficient, indicating the isometric growth when equal to 3 (Anderson and Neumann, 1996). Parameters a and b were estimated by transforming (ln) the equation by linear regression. To check the theoretical isometric (b = 3) or allometric growth (b ≠ 3), Student's t-test (Snedecor and Cochran, 1967) was employed.
The relationships between otolith and fish sizes were determined using a least-squares linear regres-sion for the following parameters: otolith length (OL) -fish length (SL) and otolith width (OW) -fish length (SL). These equations were first calculated for both left and right otoliths and the t-test was used to check any difference between regressions. The regression coefficients were compared and when significant differences (p<0.05) were not found, the H 0 hypothesis (b rigth = b left ) was accepted. When the equations did not differ statistically, a single linear regression was reported for each parameter (OL; OW) and species, by choosing randomly one right or left otolith from each specimen. The significance of the linear regressions was verified using the F-test.

Otoliths' morphology
Representative otoliths for each species are shown in Figures 2 and 3. A brief description is also provided for the species that were not reported in the atlas published by Tuset et al. (2008), according to the terminology used by these authors: i) Microstomatidae. Lanceolated anterior region and round to peaked posterior one. Broad and pointed rostrum. Lobed dorsal margin and lobed to irregular ventral margin. Median ostial sulcus acusticus, tubular ostium and tubular straight cauda. Microstoma microstoma (Fig. 2B), spindle-shaped otolith; Nansenia oblita ( Fig. 2C), spindle-shaped to sagittiform otolith.

Data analysis
Length-weight relationships and linear regression of otolith length and width against fish length for every species are given in Table 2. Of a total of 16 species, 13 showed an isometric growth since the regression coefficient b did not differ significantly from the theoretical value of 3. Statistical analysis (t-test) revealed significant differences only for three species, indicating an allometric growth: Hygophum benoiti (n = 288, df = 286, p<0.01), H. hygomii (n = 45, df = 43, p<0.001), and Vinciguerria attenuata (n = 136, df = 134, p<0.001). High correlation coefficients (R 2 >0.93) calculated for the SL-W relationship were obtained in 15 species, with the exception of Chauliodus sloani, which showed a lower value (R 2 = 0.833). In the analysis of morphometric parameters (otolith length and width) against SL, no considerable differences between right and left otoliths were detected by the t-test, so a single linear regression was plotted for each parameter. Data fitted well to the regression model for both OL and OW to SL for each species, as demonstrated by the high values of the coefficient of determination. Only the half-naked hatchetfish Argyropelecus hemigymnus showed a R 2 value that was considerably lower (R 2 = 0.813) than the others.

DISCUSSION
Otoliths are a powerful taxonomic feature for fish species identification because of their high inter-specific variability in shape. Though the classification of fish preys is facilitated by some reference works (Smale et al., 1995;Campana, 2004;Lombarte et al., 2006;Tuset et al., 2008), only certain geographical areas are covered and the access to reference material remains essential (Santos et al., 2001). Therefore, a fundamental objective of researchers studying the marine predators' feeding habits is to fill the gap of information on the fish otolith morphology and on the estimation of specific equations, which is useful to calculate the size and mass of preys.
The results of the present study respond to this need, providing SL-W, OL-SL and OW-SL relationships for several mesopelagic and bathypelagic species. Despite the importance of these fishes in top predators' diet, their biology and ecology had not been well investigated until today. In fact, owing to their null commercial value, they are not a target of fishing activities and they can only be sampled by organizing appropriate scientific cruises. The opportunity of collecting stranded specimens in the area of Strait of Messina allowed us to study these species at a low cost.
This study provides a useful tool for better understanding the trophic relationships in the Mediterranean food web. The rebuilding of prey biomass from otolith size may help to return the real importance to the Mediterranean mid-water fishes, shedding light on their role in the trophic structure, as pointed out by some feeding studies (Castriota et al., 2007;Falautano et al., 2007;Consoli et al., 2008;Karakulak et al., 2009). However, to this day, the lack of data in these area has not always permitted an appropriate quantification of their prey biomass and classification to species level in the diet of Mediterranean pelagic predators. In fact, quite a high percentage of unidentified mesopelagic fishes and otoliths in swordfish and tuna stomach contents have been recorded (Sinopoli et al., 2004;Falautano et al., 2007;Mostarda et al., 2007;Romeo et al., 2009).
In comparison with other similar studies on the relationship between fish and sagitta sizes (Wyllie Echeverria, 1987;Gamboa, 1991;Granadeiro and Silva, 2000;Harvey et al., 2000;Waessle et al., 2003), this paper supplies additional information by considering both the otolith width (OW) and length (OL). In most cases, the calculation of two equations (OL-SL and OW-SL) is more suitable, since the tip of the otolith rostrum may be damaged, making it impossible to measure the OL. Moreover, the coefficient of determination of the OW-SL linear regression attained a higher value than in the OL-SL one in the same species (Table 2). The otoliths of the investigated species did not show significant differences in size between left and right sagittae, in contrast with the findings of Waessle et al. (2003) in two sciaenid fishes (Micropogonias furnieri and Macrodon ancylodon) and Harvey et al. (2000) in the teleost Lycodes palearis (Zoarcidae). However, Harvey et al. (2000) indicated the small size of the sample as the possible cause of this diversity.
Most published regressions of OL to SL (Wyllie Echeverria, 1987;Gamboa, 1991;Harvey et al., 2000;Waessle et al., 2003) or SL to W relationships (Harvey et al., 2000;Valle et al., 2003;İlkyaz et al., 2008;Mata et al., 2008) concern coastal or commercially important species. Very few data are available on mesopelagic and bathypelagic fishes. As regards the species considered in the present paper, the sternoptychid Maurolicus muelleri has been the most investigated and SL-W and OL-SL relationships for populations living in the Norwegian Sea have been calculated (Rasmussen and Giske, 1994;Salvanes and Stockley, 1996;Kristoffersen, 2007). SL-W relationships for Argyropelecus hemigymnus and Hygophum benoiti by specimens collected in the Strait of Messina were provided by Donato et al. (1993) and Potoschi et al. (2003) respectively.
Though all data fitted well with the linear regression (OL-SL and OW-SL) and the potential (SL-W) models, it is advisable to use these equations within the fish size range limits reported in Table 1. In fact, otoliths may slow down the increase in length, increasing only in thickness at the maximum fish size (Williams and Bedford, 1974) or may grow following a curvilinear model in juvenile fishes (Nishimura and Yamada, 1988). In particular, Linkowski (1991) described a change in the otolith growth pattern at a larval size of 12 mm in the myctophid Ceratoscopelus maderensis and then supplied two different OL-SL regressions for specimens smaller than 12 mm and for larger ones. Since the individuals of C. maderensis collected in the present paper belong to the 14.5-68.7 mm SL range, the single OL-SL regression reported in Table 2 can be accepted.
Finally, some limitations to the use of biomass reconstruction from otolith size should be taken into account. The growth of individuals belonging to the same species may show some variations for different areas and stocks (Campana and Casselman, 1993;Reichenbacher et al., 2009) or between sexes (Wyllie Echeverria, 1987). Furthermore, otoliths are exposed to chemical and mechanical abrasion in the digestive track of predators (Jobling and Breiby 1986;Granadeiro and Silva, 2000) and their size may be underestimated.