Labropoulou et al,: Habitat selection and diet of luvenile Pagrus pagrus 



497 



cause all trawls were carried out with the RV Philia, 

 it was assumed that gear selectivity was constant. 

 Bottom temperature and salinity were measured 

 with a SEA-BIRD CTD unit. The measurements re- 

 corded for the 5 m above the bottom of each station 

 were averaged and the resulting values were consid- 

 ered as the temperature and salinity at the specific 

 station. 



Laboratory methods 



Fork lengths (FL, mm) were measured for at least 

 100 individuals collected at each station and cruise. 

 Of a total of 3280 fish, all specimens were found to 

 be immature. To evaluate size-related variations in 

 food habits and habitat use, red porgy were sepa- 

 rated into five size classes: (40-104 mm FL), 1 ( 105- 

 147 mm FL), 2 (148-186 mm FL), 3 (187-219 mm 

 FL), and 4 (>219 mm FL) which were chosen as they 

 approximate the age of the fish within each size class 

 as determined from the von Bertalanffy growth equa- 

 tion derived by Machias et al. ( 1998). 



Each month 20-30 specimens were preserved in 

 lO'^ buffered formalin immediately after capture for 

 stomach-content analysis. Samples were taken to the 

 laboratory, measured to the nearest mm (FL), and 

 weighed to the nearest 0.1 g. Thereafter, the stom- 

 achs were removed and the contents wet weighed. 

 Prey items were identified to the lowest possible taxo- 

 nomic level, counted under a binocular microscope 

 and weighed to the nearest 0.01 g. Prey species were 

 measured to the nearest 0.1 mm by using an ocular 

 micrometer or a vernier caliper, where possible. Fork 

 length of the fish examined ranged from 41 to 185 

 mm, mean (±SD) FL was 87.2 (±26.9) mm. 



Data analysis 



In accordance with the procedure described by 

 Carrothers (1980), the door spread of the trawl net 

 was calculated for each haul and then multiplied by 

 the speed of the boat and fishing time to estimate 

 the total area sampled. The abundance offish caught 

 was expressed as number of individuals per square 

 nautical mile (nmi'-). making the comparison offish 

 abundance between sampling stations possible. Af- 

 ter logarithmic transformations (Middleton and 

 Musick, 1986; Stefanescu et al., 1992), mean abun- 

 dances (number/nmi') were calculated for each 

 cruise: 1 ) for 20-m depth intervals, 2) for 50-m depth 

 intervals, and 3 ) per zone ( I, II, III ), as well as for each 

 1°C temperature interval. Salinity showed little vaiia- 

 tion (Table 1), having no effect on the distribution of 

 fishes on the Cretan shelf (Tsimenides et al., 1991). 



Analysis of variance showed no significant differ- 

 ences in the mean abundance among the three sur- 

 veys in each season, nor among each zone during the 

 same season (0.377<P<0.610, Bartlet test 0.201 

 <P<0.462). Results and conclusions for each season 

 were similar and independent of whether survey- 

 specific or pool data were used. Hence, only the lat- 

 ter are presented. Correlation analyses (Pearson cor- 

 relation coefficient) were performed to determine if 

 there were any significant changes in abundance 

 with depth and water temperature, as well as to test 

 the hypothesis that fish size was depth or tempera- 

 ture dependent. Geometric mean was preferred for 

 calculating the mean fish size of each sample, be- 

 cause the arithmetic mean is susceptible to the in- 

 fluence of few large specimens and does not accu- 

 rately represent the central tendency in fish size at 

 a given station (Stefanescu et al., 1992). The bathy- 

 metric distribution of different size fish was simpli- 

 fied by converting lengths into the above-mentioned 

 five size classes. 



The contribution of prey items to diet was esti- 

 mated by using both the numerical abundance and 

 the biomass of prey items in stomachs (Berg, 1979; 

 Hyslop, 1980; Bowen, 1983 ). Breadth of diet was cal- 

 culated by using Levin's standardized index 

 (Hurlbert, 1978; Krebs. 1989): 



B. 



1 



-1 



where B^ = Levin's standardized index for predator/; 

 p^ = proportion of diet of predator / that is 



made up of preyy; and 

 /; = number of prey categories. 



This index ranges from to 1; low values indicated 

 diets dominated by few prey items (specialist preda- 

 tors ), higher values indicated generalist diets (Gibson 

 and Ezzi, 1987; Krebs, 1989). Diet overlap between 

 the size classes was calculated by using the simpli- 

 fied Morisita's index (Ki-ebs, 1989; Hall et al., 1990); 



C,, = 





where C,^, = simplified Morisita's index for preda- 

 tors i and k; and 



p and Pf, = proportions of predator / and k with 

 prey 7 in their stomachs. 



Diet overlap increases as the Morisita's index in- 

 creases from to 1. Overlap is generally considered 



