4 
Fishery Bulletin 1 14(1) 
data sets, although we recognize that they would likely 
be important to consider should a larger data set be- 
come available (Venables and Dichtmont, 2004). 
For each delta-GLM, forward selection was used to 
select separately the binomial model based on presence 
and absence data and the lognormal model based on 
log-transformed data for elasmobranch CPUE. The ef- 
fects that explained more than 5% of the deviance were 
considered to have a high explanatory power (Tascheri 
et ah, 2010). Chi-square tests were run for the binomi- 
al model, and F-tests were run for the lognormal mod- 
el. The performance of the models was also compared 
by using Akaike information criterion (AIC). Analyses 
were conducted with R, vers. 3.0.2 (R Core Team, 2013). 
Patterns in species richness related to depth were 
explored with a nonparametric test, Spearman’s rank 
correlation coefficient (p), because data did not con- 
form to a normal distribution. In this analysis, the 
independent variable was depth and the dependent 
variable was the average number of species per trawl 
haul (a=0.05). Data from all survey methods conducted 
between 2010 and 2012 were pooled for this analysis. 
General patterns of sex and size segregation were an- 
alyzed in relation to depth and diel period by using 
pooled data from all types of survey conducted between 
2010 and 2012. This analysis was undertaken to deter- 
mine whether a larger proportion of females and im- 
mature individuals was caught in shallow waters. Diel 
period was included in this analysis to detect changes 
in activity levels associated with sex and maturity 
stage. Variations in the proportion of females and im- 
mature individuals with depth range (<50 m, 50-100 
m, >100 m) and diel period (day and night) were ex- 
amined with a binomial GLM (logit link) (Venables and 
Ripley, 2002). Only species with at least 100 individu- 
als were used in this analysis. 
Elasmobranch assemblage 
The importance of depth on elasmobranch assem- 
blage was further explored with PRIMER 2 , vers. 6. 2.1 
(PRIMER-E Ltd., Plymouth, UK). We used data from 
all surveys conducted in the central Pacific region dur- 
ing 2010-2012. A matrix was constructed with the 
transformed species CPUE per haul in columns (log 
[(individuals/hour )+l] ) and depth ranges in rows. To 
reduce the influence of extremely abundant species, 
CPUE was transformed (Clarke and Warwick, 2001). 
Rare species caught in less than 5 trawl hauls were ex- 
cluded from our analyses (Clarke and Warwick, 2001). 
Differences in elasmobranch assemblages among depth 
ranges were examined by using an analysis of similar- 
ity (ANOSIM; a=0.05; Clarke and Warwick, 2001). A 
similarity percentage (SIMPER) analysis was used to 
identify species that showed the highest contribution 
2 Mention of trade names or commercial companies is for iden- 
tification purposes only and does not imply endorsement by 
the National Marine Fisheries Service, NOAA. 
to the dissimilarities among depth ranges (Clarke and 
Warwick, 2001). 
Redundancy analyses (RDAs) were applied to exam- 
ine the relationship between environmental variables 
and elasmobranch assemblages (Borcard et al., 2011). 
An RDA is a constrained ordination technique that com- 
bines a multivariate, multiple linear regression with a 
principal component analysis. We performed RDAs that 
were based on covariance matrixes to confer a higher 
weight to the common species in these analyses. 
A separate RDA was conducted for each survey 
type to avoid biases that may have resulted from 
combining the 3 survey methods. For all analyses, a 
Hellinger transformation was applied to the species 
CPUE to minimize the effects of the large number of 
zeros in the data set (Borcard et al., 2011). Rare spe- 
cies (caught in less than 5 trawl hauls) were excluded 
from the analysis to prevent strong distorting effects. 
The environmental variables considered in the RDAs 
were standardized depth, standardized latitude, year, 
season, and diel period. The statistical significance of 
the ordination axes was examined with a Monte Carlo 
permutation test. Results were plotted on a correla- 
tion biplot, in which angles between species and en- 
vironmental variables represent correlations between 
variables (Borcard et al., 2011). These analyses were 
conducted by using the vegan library in R, vers. 3.0.2 
(R Core Team, 2013). 
Results 
For this study, data were examined from 346 trawl 
hauls conducted along the entire Pacific coast of Cos- 
ta Rica from 2008 to 2012. Of these hauls, 108 were 
from deep water surveys, 111 were from monitor- 
ing surveys, and 127 were from commercial surveys 
(Fig. 1). Commercial and monitoring sampling efforts 
were highest in the central Pacific region, where 76% 
of commercial hauls and 91% of monitoring surveys 
were conducted (Table 1). Most commercial trawl 
hauls occurred at depths <100 m, and the majority of 
monitoring and all deepwater trawl hauls were con- 
ducted in deeper waters (Table 1). The average stan- 
dardized elasmobranch abundance was 9.37 individu- 
als/hour in deep water surveys, 6.96 individuals/hour 
in monitoring surveys, and 7.92 individuals/hour in 
commercial surveys. 
Elasmobranch diversity and distribution patterns 
During the entire sampling period, 4564 elasmobranchs 
from 25 species, 13 families, and 6 orders were cap- 
tured as bycatch (Table 2). Four species represented 
more than 66% of the entire elasmobranch abundance: 
Panamic stingray (Urotrygon aspidura ) accounted for 
26%, rasptail skate (Raja velezi) contributed 16%, 
brown smoothhound (. Mustelus henlei) composed 15%, 
and witch guitarfish ( Zapteryx xyster ) accounted for 
9%. Of the remaining 21 species, 10 were relatively 
