56 
Fishery Bulletin 119(1) 
standardized index (Levins, 1968). This index ranges from 
0 to 1, with values close to 0 indicating a specialized diet 
and those close to 1 indicating more generalized feeding 
habits. This index was computed as follows: 
B. = pee x a nae (2) 
1 ? 
N-1 Deitel 
where B; = Levins’s standardized index; 
pi; = the proportion of prey j in the diet of predator 
1; and 
N = the total number of prey groups. 
The trophic niche overlap among the different studied 
species was estimated by using the simplified Morisita 
index (Morisita, 1959), which compares pairs of species 
with values ranging between 0.00 (no prey overlap) to 1.00 
(full prey overlap) as follows: 
C 2y PiPi« 
Je 
» > pa+ > pe 
where C;,, = the simplified Morisita index for predators i 
and k, and 
pj; and p,,; = the proportions of predators i and k with prey 
J in their stomachs. 
(3) 
The trophic level (T) was calculated by using the fol- 
lowing formula developed by Christensen and Pauly 
(1992): 
where DC;; = the proportion of prey j in the diet of the 
predator i, and 
NT, = the trophic level of prey j. 
Trophic levels of prey were determined empirically by 
using local information or, when this information was 
not available, by using the modeled area (Cortés, 1999; 
Ebert and Bizzarro, 2007; Fanelli et al., 2011; Jacobsen 
and Bennett, 2013; Corrales et al., 2015; for details 
on the species of prey for which these cited references 
were sources of local information on trophic levels, see 
Supplementary Table 2 [online only]). In this study, the 
trophic level of each of the defined categories based on 
species and size class was first calculated by weighting 
its average biomass obtained during the surveys of the 
MEDITS program conducted in 2011-2017. 
Co-occurrence measures 
To analyze patterns of species associations related to 
food resource partitioning, we studied interspecific and 
intraspecific co-occurrence. Therefore, for the entire study 
period, we accounted for the abundance of all species and 
size classes estimated during surveys. The abundance 
of Mediterranean bigeye rockling was underestimated 
because of the low selectivity of the gear used to catch this 
small species. Therefore, this species was not included in 
the co-occurrence analysis. The Jaccard index (S;;) was 
used to measure co-occurrence, as an expression of associ- 
ation between species (Jaccard, 1901): 
S;=a/(a+b-+e), (5) 
where a = the number of occasions in which both species or 
size classes are present; 
b = the number of occasions in which only one of the 
species is present; and 
c = the number of occasions in which only the other 
species is present. 
This index was calculated as the percentage of occasions 
that both species or size classes appeared in the same haul. 
In this study, the only hauls that were considered were 
those in which at least one of the species or size classes of 
each pair was present. 
Statistical analyses 
All data were standardized by using square-root transfor- 
mation, and a Bray—Curtis similarity matrix was calcu- 
lated. One-way analyses of similarity were performed to 
identify significant differences (P<0.05) in trophic strat- 
egies between pairs of size classes (i.e., intraspecific com- 
petition). We used the software PRIMER 6 (PRIMER-e, 
Quest Research Ltd., Aukland, New Zealand) (Clarke and 
Gorley, 2006) to estimate global R as a scaled measure of 
the separation between groups of samples, with values 
ranging from 0 (no differences) to 1 (completely different) 
(Clarke, 1993). 
Similarity percentage analysis was applied to determine 
which prey groups contributed most to the dissimilarity in 
diet composition between such pairs. According to these 
results, new species-and-size categories were considered 
for further analyses. Finally, clustering methods and mul- 
tidimensional scaling were used to analyze prey affinities 
and to discern feeding strategies that possibly were dif- 
ferent between the studied species among size classes. All 
calculations were done by using PRIMER 6. 
Results 
Diet composition and ontogenetic variation 
Overall, for all studied species as a group, the most common 
prey groups in terms of %V were large plankton; crab, lob- 
ster, and mantis shrimp species; shrimp species; and ben- 
thopelagic fish species. Large plankton, represented mainly 
by euphausiids, was the most common prey (mostly for sil- 
very pout and blue whiting). In the case of decapod species, 
the red snapping shrimp (Alpheus glaber), green shrimp 
(Chlorotocus crassicornis), and angular crab (Goneplax 
rhomboides), all members of Pleocyemata, were the most 
abundant prey (mostly for Mediterranean bigeye rockling, 
poor cod, and greater forkbeard). Finally, benthopelagic fish 
species were primarily represented by blue whiting and 
myctophids as prey for studied species (mostly for Mediter- 
ranean ling and blue whiting) (Table 3). The raw data set 
