312 
Fishery Bulletin 116(3-4) 
timateS, vers. 9.1.0 (Colwell, 2013). The order of sam¬ 
ples was permuted 1000 times to reduce bias. Follow¬ 
ing the method of Jimenez-Valverde and Hortal (2003), 
we plotted the results to analyze the suitability of the 
sample size. When the curve approaches the asymp¬ 
tote (slope<0.1), the number of samples is deemed to 
be sufficient to describe the diet (Soberon and Llorente, 
1993). Furthermore, we used the calculation method 
proposed by Bizzarre et al. (2007) to reinforce the sam¬ 
pling assessment. This method states that the slope 
of the line generated from the curve endpoints (mean 
cumulative number of prey taxa generated for the final 
4 stomach samples) should be compared to a line of 0 
slope to establish whether a cumulative prey curve has 
reached an asymptote. Slopes are compared by using 
Student’s t-test, where slopes that are not significant 
(P>0.05) indicate that the curve has reached an asymp¬ 
tote (Bizzarro et al., 2007). 
The relative importance of each prey species to the 
diet of the blue shark was established by the prey-spe¬ 
cific index of relative importance (%PSIRI) (Brown et 
al., 2012), by using the equation: 
%PS IRI-0.5 %FO , x ( %PN\+%PW^, (1) 
where %FO { = the number of stomachs containing prey 
category i, divided by the total number 
of stomachs n; 
%PNi = prey-specific numeric abundance; and 
%PWi - prey-specific wet-weight abundance. 
Prey-specific abundance ( %PA X ) was calculated by 
%PA i = 'Ll 1 %A ij n;\ 
where = the abundance (by counts [%PN^\ or weight 
[%PWj]) of prey category i in stomach 
sample j; and 
tti = the number of stomachs containing prey i. 
The %PSIRI, is a modification of the index of relative 
importance (%IRI [Cortes, 1997]) that avoids %FO re¬ 
dundancies taken in the %IRI and is additive with re¬ 
spect to taxonomic levels. As a result, the %PSIRI of a 
family will be equal to the sum of the %PSIRI of the 
species within the taxon (Brown et al., 2012). 
Niche breadth was calculated according to Levin’s 
standardized index by using %PSIRI converted to pro¬ 
portions at the family level (Krebs, 1999). The index 
values ranged between 0 and 1, where values closer to 
0 indicate a diet dominated by few prey species (i.e., by 
a greater degree of specialization) and values closer to 
1 indicate a lesser degree of specialization (Munroe et 
al., 2014). In addition, the graphical analysis proposed 
by Amundsen et al. (1996) was performed to explore 
prey importance at the family level and predator feed¬ 
ing strategy. The analysis is based on a 2-dimensional 
graph representation of prey-specific abundance (%PA ; ) 
in relation to the frequency of occurrence of the differ¬ 
ent prey types in the diet (%FO ; ). 
Trophic position was calculated based on percent 
weight values of the prey species identified with the 
equation proposed by Christensen and Pauly (1992): 
TP = 1 + (X? =1 DC y )x(7^.), (2) 
whereUCij = the composition of the prey j in the diet of 
the predator /; 
TPj = to the trophic level of prey j; and 
n = the number of prey species in the diet of 
predator j. 
Values of trophic position for fish prey were obtained 
from Froese and Pauly 3 and Espinoza (2014); and 
for cephalopod prey from Cortes (1999) and Espinoza 
(2014). 
To identify possible differences in diet, individual 
sharks were analyzed according to 5 factors: sex, sea¬ 
son, size class, latitude of fishing ground, and longitude 
of fishing ground. The analysis by sex was performed 
to clarify whether the composition of diets of females 
and males is related to the reported spatial segregation 
by sex (Nakano and Stevens, 2008). Specimens ana¬ 
lyzed per season were grouped into warm (February- 
May and December) and cold (June-November) seasons 
according to established patterns of sea-surface tem¬ 
peratures (SST) (Flores et al., 2013). Cluster analysis 
was employed with 20-, 30-, and 40-cm-TL intervals 
to define size classes (Markaida and Sosa-Nishizaki, 
2010) by using the numeric abundance (at family lev¬ 
els) of identified preys. Calculations were performed 
by the unweighted pair-group method with arithmetic 
mean and by using the Bray-Curtis index as a mea¬ 
sure of dissimilarity. A 50% of dissimilarity distance 
indicated major divisions between size classes (Ebert 
and Bizzarro, 2007). Fishing grounds based on latitude 
were grouped into 2 biogeographical marine provinces, 
where ‘northern’ corresponds to the Tropical Eastern 
Pacific marine province and ‘Northern-Central’ to the 
Warm Temperate Southeastern Pacific marine province 
(Spalding et al., 2007). Fishing grounds based on lon¬ 
gitude were grouped into ‘coastal’ and ‘oceanic’ groups, 
with the Peru-Chile Trench (which occurs at an aver¬ 
age distance from the coast of 130 km or 70 nautical 
miles) as the boundary between these 2 groups (Mach- 
are et al., 1986) (Fig. 1). 
Statistical analysis 
To assess differences in the diet of blue sharks by fac¬ 
tors, we performed two multivariate techniques: non¬ 
metric multidimensional scaling (MDS) ordinations 
and analysis of similarity (ANOSIM). These techniques 
were conducted with the Bray-Curtis index of dissimi¬ 
larity generated from the numeric abundance of each 
prey grouped by family (Mendoza-Avila et al., 2016), 
pretreated by fourth-root transformation and standard¬ 
ized to percentages. The stress value generated by the 
nonmetric MDS model indicates the reliability of the 
representation, where values closer to 0 indicate excel¬ 
lent representation and values larger than 0.2 indicate 
that interpretation of the data is unreliable (Clarke, 
3 Froese, R., and D. Pauly (eds.). 2018. FishBase, vers. 
02/2018. World Wide Web electronic publication. [Avail¬ 
able from website.] 
