374 
Fishery Bulletin 111(4) 
Chessel, 1994; Dray et al., 2003). This method has been 
used successfully on diverse ecological data sets and 
organisms, including fishes (e.g., Mellin et ah, 2007; 
Carassou et al., 2011; Lecchini et ah, 2012), zooplank- 
ton (e.g., Carassou et ah, 2010), benthic invertebrates 
(e.g., Bremner et ah, 2003), and bacteria (e.g., Jardillier 
et ah, 2004). In our study, each COIA was based on 
the matching of a normed PCA of environmental data 
and a centered PCA of shark abundance data (PCA- 
PCA-COIA, Dray et ah, 2003). Monte Carlo tests with 
10,000 permutations between observations were used 
to confirm the significance of COIA results (fixed-D 
test; Dray et ah, 2003), with significance assessed at 
P<0.05. For each COIA, the vectorial correlation (RV) 
coefficient, a multivariate generalization of the squared 
Pearson’s correlation coefficient, provided a quantita- 
tive measure of the co-structure between explanative 
(environmental) and explained (shark CPUE) vari- 
ables, with a value of 1 indicating a perfect match be- 
tween the 2 data sets (Doledec and Chessel, 1994; Dray 
et ah, 2003). The criterion of total inertia was used to 
compare the amount of agreement between environ- 
mental and shark data for the 2 spatial scales consid- 
ered (Dray et ah, 2003). All multivariate analyses were 
performed with the ADE-4 software (Thioulouse et ah, 
1997, 1995-2000). 
Results 
Small-scale sampling 
During small-scale sampling, 353 stations were sur- 
veyed, spanning the months from March to November 
during 2006-09 (Fig. IB). Winter months (December, 
January, and February) were excluded from subsequent 
analyses because of the complete absence of sharks in 
the small-scale survey area during this time (2100 
hooks with no sharks). Over the course of this survey, 
2417 individuals representing 12 shark species were 
encountered. Of these 12 species, 5 species met our 
criteria for inclusion in subsequent analyses (i.e., they 
also were abundant in the large-scale data set): At- 
lantic Sharpnose Shark ( Rhizoprionodon terraenovae), 
Blacktip Shark ( Carcliarhinus limbatus), Blacknose 
Shark (C. acronotus ), Spinner Shark (C. brevipinna), 
and Bull Shark (C. leucas). Mean CPUE (±standard er- 
ror [SE] ) ranged from 2.88 [0.28] sharks 100 hooks -1 h -1 
for Atlantic Sharpnose Shark to 0.11 (0.02) sharks 100 
hooks -1 h -1 for Bull Shark (Table 1). Wide size ranges, 
with size measured as fork length (FL) in centimeters, 
were found for Atlantic Sharpnose (36.0-96.3 cm FL), 
Blacktip (59.8-164.0 cm FL), Blacknose (40.9-136.0 cm 
FL), and Spinner (49.9-165.9 cm FL) Sharks. A small- 
er size range was seen for Bull Sharks (73.0-155.5 cm 
FL), the least commonly encountered of the 5 species 
(Table 1). 
The centered PCA conducted with small-scale data 
on shark abundance explained 91.88% of the variabil- 
ity between observations (across blocks 1-8) on the 
first 2 principal components (PCI and PC2) (Fig. 3A). 
Variation along PCI was explained primarily by data 
for Atlantic Sharpnose Shark, which was most abun- 
dant in blocks 2, 3, and 4 (western blocks), less com- 
mon in block 1, and relatively rare in blocks 5, 6, 7, and 
8 (eastern blocks). Spinner Shark showed a similar but 
less marked spatial pattern (Fig. 3A). Variation along 
PC2 was explained primarily by data for Blacktip 
Shark, which was more abundant in block 1 (western 
block), and relatively rare in block 5 and 6 (eastern 
blocks) (Fig. 3A). Patterns were less clear for Blacknose 
and Bull Shark. 
The normed PCA on small-scale environmental data 
explained 74.18% of the variability between observa- 
tions (blocks) on the first 2 principle components (PCI 
and PC2) (Fig. 4A). Temperature and crustacean bio- 
mass were positively correlated with each other and 
both of those variables had a high negative correlation 
with salinity. These 3 variables explained most of the 
variability along PCI. Fish biomass was negatively cor- 
related with depth. Chl-a concentration and dissolved 
oxygen were negatively correlated, together explain- 
ing most of the variability along PC2. Blocks 7 and 8 
(eastern blocks) were characterized by high dissolved 
oxygen and low concentration of chl-a, and the inverse 
was true for block 3 (a western block) (Fig. 4A). 
The COIA that coupled small-scale shark abundance 
and environmental data was characterized by a total 
inertia of 0.22 and an RV coefficient of 0.65, indicat- 
ing a relatively high degree of agreement between the 
structures of the 2 data sets. Axes 1 and 2 supported 
99.17% of this common structure (Fig. 5A). Atlantic 
Sharpnose Shark abundance was positively related 
with chl-a concentration and negatively related with 
dissolved oxygen and salinity. Abundance of Blacktip 
Shark was more positively associated with crustacean 
biomass than wfith other environmental variables. 
Blacknose and Spinner Sharks had high negative 
associations with dissolved oxygen, and Blacknose 
Shark had a strong positive association with depth 
(Fig. 5A). 
Large-scale sampling 
Across the large-scale survey area, shark abundance 
data were obtained from 551 stations sampled during 
the months of August and September during 2006-09 
(Fig. 2B). Over the course of this survey, 4493 sharks, 
comprising 26 species, were captured. Mean catch per 
unit of effort (±SE) ranged from 4.74 (0.41) sharks 
100 hooks -1 h -1 for Atlantic Sharpnose Shark to 0.06 
(0.01) sharks 100 hooks -1 h -1 for Bull Shark (Table 1). 
Wide size ranges were observed for Atlantic Sharpnose 
(33.0-115.5 cm FL), Blacktip (38.2-157.0 cm FL), Blac- 
knose (40.0-104.9 cm FL), and Spinner (54.0-169.0 cm 
FL) Sharks. The smallest size range was seen in Bull 
Shark (131.4-176.0 cm FL), the least commonly en- 
countered of the 5 species (Table 1). 
