Clarke et al.: Elasmobranch bycatch from the shrimp trawl fishery along the Pacific coast of Costa Rica 
3 
Figure 1 
Map of the geographic regions and sampling locations for elasmo- 
branch bycatch along the Pacific coast of Costa Rica, Central America, 
during 2010-2012. Solid lines represent the 50-m, 200-m, and 500-m 
depth contours. Dotted lines represent the boundaries of the central 
Pacific region. 
and February 2011). These surveys followed a system- 
atic sampling design, in which 15-min trawl hauls were 
conducted at 3 different depths: 150, 250, and 350 m. 
Hauls were conducted in areas where shrimp were 
expected to be caught. Strict grids were not used to 
determine sampling sites in order to respect marine 
protected areas but were distributed as evenly as pos- 
sible along the coast. 
Monitoring surveys were part of a program designed 
to analyze crustacean bycatch and were carried out on 
a monthly basis between 2010 and 2012; they consisted 
of one nocturnal and one diurnal set of four 20-min 
trawl hauls conducted at depths of approximately 100, 
140, 180, and 220 m (Fig. 1). The location of each haul 
was determined by the vessel’s captain; therefore, the 
majority of the sampling effort was concentrated in 
shrimp fishing grounds in the central Pacific region 
(Fig. 1). These sampling stations were chosen because 
of their general proximity to the main port of Pun- 
tarenas and their high probability of yielding large 
catches of shrimps, according to the captain’s previous 
experience. 
Commercial sampling was carried out during the 
same trips as those conducted by the monitoring sur- 
veys. Commercial sampling points were not selected on 
the basis of a systematic grid; instead, sampling oc- 
curred at locations where the captain had previously 
targeted shrimps. Sampling occurred on a monthly 
basis from April 2010 to August 2012 and includ- 
ed trawl hauls conducted at depths of 18- 
350 m. 
Sampling for all 3 surveys was carried out 
aboard commercial shrimp trawlers (22.5 m), 
equipped with a 270-hp engine and 2 standard 
epibenthic nets (20.5 m long; mouth opening 
of 5.35x0.85 m; mesh size of 4.45 cm; and co- 
dend mesh size of 3.0 cm). Trawl speed varied 
between 2.1 and 5.7 km/h during all surveys. 
Information recorded for each trawl haul in- 
cluded geographic coordinates (latitude and 
longitude), depth (measured in meters with an 
installed sonar), and trawl duration (defined 
as the time, in minutes, during which the net 
was on the bottom). 
Elasmobranchs were identified, classified 
according to sex, measured (total length [TL] 
for sharks and disc width [DW] for rays), and 
weighed (total weight [TW]) (Bussing and 
Lopez, 1993; Compagno et al., 2005). Maturity 
stage was assessed by macroscopic examina- 
tion of the reproductive tract (Conrath, 2005; 
Clarke et al., 2014). 
General abundance and distribution patterns 
The effects of depth, latitude, year, diel pe- 
riod (day: 0600-1800; night: 1800-0600), 
and season (rainy and dry) on elasmobranch 
abundance were examined by using a delta- 
lognormal generalized linear model (delta- 
GLM). This method is commonly applied to zero-inflat- 
ed fishery data, which tend to violate key assumptions 
of many statistical techniques (Stefansson, 1996). The 
delta-GLM approach comprised 2 stages: 1) elasmo- 
branch presence and absence data were modeled by 
using a binomial GLM with a logit-link, and 2) the 
observed positive densities were modeled with a log- 
transformed positive subset, which was assumed to 
be Gaussian with an identity link function. Because 
of the differences in the sampling design between the 
3 survey methods, separate delta-GLMs were applied 
to deepwater, monitoring, and commercial data. Total 
elasmobranch abundance was standardized to catch 
per unit of effort (CPUE), defined as the number of 
individuals per hour of trawling. 
For deepwater surveys, the independent variables 
considered in the analyses were depth, latitude, year, 
and season. Diel period was excluded from the model 
for deepwater trips because hauls were carried out only 
during the day. The independent variables considered 
in models for monitoring and commercial surveys in- 
cluded depth, latitude, year, season, and diel period. In 
all 3 models, depth and latitude were treated as contin- 
uous variables, and year, diel period, and season were 
treated as factors. In order to avoid strong interactions 
between depth and longitude, depth and latitude were 
used to represent the geographic location of each trawl 
survey. Interactions between variables could not be 
considered because of the small size of the available 
