Bradford et al.: Abundance estimates of cetaceans within the U.S. Hawaiian Islands EEZ 
131 
The observation methods used during the HICEAS 
in 2010 were developed by the SWFSC and have been 
in use for the last 3 decades (e.g., Barlow, 2006). To 
summarize these methods, observation teams consisted 
of 6 observers who rotated through 3 roles (port and 
starboard observers and a data recorder) and searched 
for cetaceans 180° forward of the vessel by using 25x 
binoculars (port and starboard observers) and with un¬ 
aided eyes (data recorder) from the flying bridge (ap¬ 
proximately 15 m above the sea surface on both ships). 
When cetaceans were sighted within 5.6 km (3 nmi) 
of the trackline by 1 of the 3 on-effort observers, sys¬ 
tematic search effort was suspended and the ship di¬ 
verted from the trackline toward the sighting so that 
species, species composition (for mixed-species groups), 
and group size could be determined. In addition to ba¬ 
sic environmental data (e.g., Beaufort sea state, swell 
height, and visibility), data collected for each sighting 
included the time, location, initial bearing and radial 
distance to the cetacean group (used to calculate the 
perpendicular distance of the sighting to the trackline), 
species identity, proportion of each species present 
(mixed-species groups), and identity of observers and 
their independent estimates of sighting group size (re¬ 
corded as a “best,” “high,” and “low” estimate for each 
observer). If species identity could not be determined 
for a sighting, the lowest possible taxonomic category 
was applied (see Table 1 for the categories relevant to 
the HICEAS in 2010). After the identification of spe¬ 
cies and estimation of group size for some sightings, 
depending on weather, animal behavior, and research 
priorities, a small boat was launched to collect photo¬ 
identification images and biopsy samples. 
Additionally, an acoustics team worked independent¬ 
ly from the observers, detecting cetacean vocalizations 
by using a hydrophone array towed behind each ship 
during daylight hours. This team did not inform the 
observer team of acoustic detections. The abundance 
estimation reported in the present study is based solely 
on the sightings made by the observers. That is, ce¬ 
taceans that were detected only acoustically were not 
included in the abundance analysis. The acoustic de¬ 
tections from the HICEAS in 2010 are currently being 
processed for future line-transect analyses. 
Estimation of abundance 
Cetacean abundance in the Hawaiian Islands EEZ was 
estimated by using a multiple-covariate line-transect 
approach (Buckland et al., 2001; Marques and Buck- 
land, 2004). Specifically, detection functions were mod¬ 
eled as a function of factors known to affect the detect¬ 
ability of cetacean groups. Sighting rates are low in 
the Hawaiian Islands EEZ (Barlow, 2006), and as were 
the sample sizes during the HICEAS in 2002, sample 
sizes for each species sighted during the HICEAS in 
2010 were inadequate for modeling the detection func¬ 
tions. Therefore, as with analysis of sightings from 
the HICEAS in 2002 (Barlow, 2006), sightings from 
the HICEAS in 2010 were pooled with sightings col¬ 
lected during previous NMFS ship-based line-transect 
surveys of the eastern Pacific. The estimation of de¬ 
tection functions for the HICEAS in 2002 incorporated 
sightings made throughout the eastern Pacific during 
SWFSC surveys conducted from 1986 through 2002, 
but the sighting pool for the analysis of the 2010 data 
was restricted to sightings made in the central Pacific 
(defined here as the area of the eastern Pacific north of 
5°S, south of 40°N, west of 120°W, and east of 175°E) 
during SWFSC and PIFSC surveys from 1986 through 
2010. The pooled sightings (collected during both sys¬ 
tematic and nonsystematic efforts) were limited to the 
central Pacific to minimize heterogeneity resulting 
from geographical differences in species associations 
and behavior—complex factors that can be difficult to 
represent as covariates. 
Despite survey data from the present study being 
pooled with previous survey data, sample sizes for 
most species remained insufficient for estimating the 
detection function. Therefore, sightings of species with 
similar detection characteristics (e.g., size, surface be¬ 
havior, group sizes) were also combined for modeling 
the detection function. Specifically, 6 species pools were 
formed: 1) small delphinids with relatively large group 
sizes; 2) small and medium delphinids with relatively 
small group sizes; 3) large delphinids and co-occurring 
beaked whales with similar behavior (Barlow, 2006); 
4) large and highly conspicuous odontocetes (Barlow 
et al., 2011a); 5) beaked whales with relatively small 
group sizes; and 6) baleen whales (see Table 2 for the 
composition of each species pool). 
A half-normal model was used to evaluate the de¬ 
tection probabilities for the sightings in each species 
pool as a function of perpendicular distance from the 
trackline and of relevant covariates. Only half-normal 
models were used because of the greater stability 
they exhibit when fitting sighting data for cetaceans 
(Gerrodette and Forcada, 2005). The 5-10% most dis¬ 
tant sightings in each species pool were truncated to 
improve model fit (Buckland et al., 2001), although 
no truncation distance exceeded the 5.6-km limit at 
which the ship would not divert from the trackline 
for a sighting. Covariate models were built by using a 
forward stepwise procedure and were selected by us¬ 
ing Akaike’s information criterion corrected for a small 
sample size (AICc; Hurvich and Tsai, 1989). 
Although several factors have the potential to af¬ 
fect the perpendicular sighting distances to cetaceans 
(Barlow et al., 2001), a smaller set of covariates identi¬ 
fied as important and robust in estimating detection 
probabilities (Barlow et al., 2011a) was considered for 
analysis in the present study. Of the covariates iden¬ 
tified by Barlow et al. (2011a), visibility and swell 
anomaly could not be tested because these variables 
were not recorded during SWFSC surveys before 1991, 
and region was not applicable because the pooled sight¬ 
ings were restricted to the central Pacific. The remain¬ 
ing covariates evaluated were Beaufort (Beaufort sea 
state, treated as a continuous variable), group size (the 
natural logarithm of the sighting group size, which in- 
