226 
Fishery Bulletin 1 14(2) 
Table 3 
Minimum (Min), maximum (Max), and mean depth ( Z a ) of the fish aggregations at 
the water column and the distance from shore for ECOSAR III-VII cruises. Stan- 
dard deviations (SDs) of the means are presented in parentheses 
Z a (m) Distance from shore (km) 
Cruise 
Min 
Max 
Mean (SD) 
Min 
Max 
Mean (SD) 
III 
7.5 
60 
24.4 (14.6) 
4.0 
113.8 
38.9(35.5) 
IV 
11 
55 
20.6(10.3) 
2.6 
51.7 
22.9 (16.6) 
V 
13 
84 
32.6 (22.4) 
2.8 
78.0 
27.7 (25.0) 
VI 
12 
34 
19.6 (6.1) 
1.0 
132.2 
38.7 (38.6) 
VII 
12 
58 
29.0 (14.7) 
4.2 
100.2 
38.5 (30.2) 
the 20 most parsimonious models according to both cri- 
teria. Then, that model was chosen that had the low- 
est BIC value among those models with the number of 
variables defined above (Anderson et al., 2008). 
The influence of the predictor variables of the se- 
lected model on the variation in fish assemblage struc- 
ture was visually assessed by the constrained ordi- 
nation distance-based redundancy analysis (dbRDA) 
(Anderson et al., 2008). Vectors on dbRDA ordination 
determine the strength and direction of the relation- 
ship between the predictors and the ordination axes. 
The length of the vectors was set proportionally to r, 
and their direction indicated the direction of the great- 
est increase in the variable. To explore the relationship 
between species and predictors, both species and pre- 
dictor variable vectors were shown on the ordination. 
Only species showing a Spearman correlation coeffi- 
cient ( | r SP | ) >0.2 (Willis and Anderson, 2003) with one 
or more axes were displayed. 
The pattern of species association We performed a clus- 
ter analysis 1) to detect potential groupings of species 
on the basis of their frequency of co-occurrence and 2) 
to explore the relation of these groupings with the sig- 
nificant predictors, plotting the groupings on the dbR- 
DA ordination. Therefore, we used the same response 
matrix as that which we used for DistLM, reducing the 
number of species to include only those species contrib- 
uting to (arbitrarily) 93% of the relative total frequency 
of occurrence. The data were transformed into presence 
and absence, and an R-mode cluster was constructed 
on the basis of a matrix of similarities with the Ochiai 
algorithm (Podani, 2000) and distances calculated by 
group-average sorting. The species groupings were ar- 
bitrarily defined at a cutoff level of 20% similarity. 
Results 
Abiotic variables 
Fish sampling sites were in areas of the SBB at depths 
from 20 to 103 m (mean per cruise: 46.5 m [standard 
deviation (SD) 20.9]). The tows were, on average, lo- 
cated from 22.9 to 38.8 km offshore, with the nearest 
tow at 1 km from the shore and the farthest tow at 
132 km (Fig. 1, Table 3). The values for Z a ranged from 
19 m to 32 m, with the shallowest aggregation at 7.5 
m (ECOSAR III cruise) and the deepest aggregation at 
84 m (ECOSAR V cruise) (Table 3). During most tows 
(60.5%), the net was hauled more than 3 m above the 
bottom (with a mean distance of 16.4 m [SD 14.4]) from 
the bottom of the net to the seafloor, and, during 6 tows 
(7.7% of all tows), the net was hauled from 1 to 3 m 
from the bottom. The net reached the bottom in 32% 
of the hauls. 
The number of fish samples taken in each water 
mass varied among the ECOSAR cruises (Fig. 2). Of 
the 78 fish samples taken during the cruises, 43% were 
collected in SACW, 32% were captured in M, and 24% 
were taken in CW (Fig. 2). 
Species composition and abundance 
In the valid samples used in this study, 43 species 
from 23 families were recorded (Table 4). Carangidae 
was the most speciose family (9 species), followed by 
Engraulidae (5 species). The Argentine anchoita ( En - 
graulis anchoita [Engraulidae]), rough scad ( Trachurus 
lathami [Carangidae]), Atlantic cutlassfish ( Trichiurus 
lepturus), Brazilian sardinella, and flying gurnard ( Dac - 
tylopterus volitans ) were the most abundant species 
(70% of the total biomass sampled). These 5 species, 
together with the Atlantic chub mackerel ( Scomber co- 
lias [Scombridae]), gray triggerfish ( Balistes capriscus 
[Balistidae]), American coastal pellona ( Pellona har- 
roweri [Pristigasteridae] ), piquitinga anchovy ( Anchoa 
tricolor [Engraulidae]), and harvestfish ( Peprilus paru 
[Stromateidae]), accounted for 91% of the total biomass 
sampled (Table 4). The predatory Atlantic cutlassfish 
was the most common species in the sampled aggrega- 
tions (Table 4). 
Large schools of Argentine anchoita and rough 
scad dominated the biomass of the aggregations and 
were moderately common during the ECOSAR cruises 
(FO t =~ 41% and 26%, respectively; Table 4). This find- 
