524 
Fishery Bulletin 107(4) 
data on the landed species. A portion of the vessels, 
randomly selected, also carry observers who record all 
catches including noncommercial species. 
Fishing effort in the Hawaii-based deep-set long- 
line fishery targeting bigeye tuna ( Thunnus obesus ) in- 
creased about 250% between 1996 and 2006. The num- 
ber of fishing sets increased from about 530 per month 
to 1370 per month, and the number of hooks deployed 
increased from about 850,000 per month to 2.9 million 
per month. The catch also increased from 161,000 to 
427,000 fishes annually between 1996 and 2006. 
In this article, changes in catch rates were investi- 
gated within the upper trophic levels of the subtropical 
ecosystem. Logbook and observer data from the Hawaii- 
based deep-set longline fishery provided catch and effort 
data that were used to describe the changes in catch 
rates of the most commonly caught commercial and 
noncommercial species from 1996 to 2006. Ecological 
indicators of the catch were also computed to estimate 
trends in the exploited ecosystem. 
Material and methods 
The Hawaii-based longline fishery consists of two com- 
ponents: the daytime deep-set fishery targeting bigeye 
tuna at depths, and the nighttime shallow-set fishery 
targeting swordfish ( Xipliias gladius ). The deep-set 
fishery typically sets hooks between depths of 100 m 
to 400 m with the median hook depth at about 250 
m (Bigelow et al., 2006). Catch data recorded by fish- 
ermen in federally mandated logbooks provide daily 
records of fishing activity such as location, catch by 
species, number of hooks per set, and since 1996, the 
number of hooks per float for each set. Deep sets and 
shallow sets can be identified by a very strong bimodal 
distribution of the number of hooks between floats. For 
shallow sets, 2-6 hooks are used per float, whereas 
for deep sets, 20-32 hooks are used per float (Bigelow 
et al., 2006). For our analysis we identified deep sets 
as those with 10 or more hooks per float and shallow 
sets as those with fewer than 10 hooks per float. The 
shallow-set fishery operates primarily in the winter 
and spring within a narrow band of 28-32°N latitude. 
The shallow-set fishery was closed for several years to 
reduce interactions with sea turtles. In this article we 
focus exclusively on the deep-set fishery that operates 
throughout the year over a broad geographic region 
and provides an uninterrupted catch and effort time 
series from 1996. The restriction of our analysis to the 
deep-set fishery provides a relatively standardized depth 
range and method of gear deployment. Our analysis was 
further restricted to data that were obtained from the 
core region of the fishing ground defined as bounded 
by 12-27°N latitude. In some years, the fishery made 
excursions as far south as the equator and as far north 
as 32°N latitude; however, fishing in these areas was 
inconsistent over the period of the study. 
In addition to logbook records of all commercially 
valuable catches, a portion of the longline vessels car- 
ried observers who recorded all catches and measured a 
subset of the catches. Between 1996 and 2006 approxi- 
mately 16% of the deep-set effort in the core fishing 
ground had observer coverage. The top 13 species in 
the catch, determined from the observer data, accounted 
for 90% by number of the total observed catch over the 
period 1996-2006 in the deep-set fishery in the core 
fishing ground. In descending order of their proportion 
in the catch they were bigeye tuna ( Thunnus obesus ), 
longnose lancetfish ( Alepisaurus ferox), blue shark ( Prio - 
Jiace glauca ), mahimahi (Coryphaena hippurus), sickle 
pomfret ( Taractichthys steindachneri), snake mackerel 
( Gempylus serpens), skipjack tuna ( Katsuwonus pelamis ), 
albacore ( Thunnus alalunga), yellowfin tuna ( Thun- 
nus albacares ), striped marlin ( Tetrapturus audax), 
escolar (Lepidocybium flavobrunneum ), ono ( Acantho - 
cybium solandri), and shortbill spearfish ( Tetrapturus 
angustirostris). The local name most frequently used 
in Hawaii for the sickle pomfret is monchong. Other 
common names used for mahimahi and ono are dolphin- 
fish and wahoo, respectively. Three species reported by 
observers as part of the catch but not fully reported 
in logbooks because of their limited commercial value 
were lancetfish, snake mackerel, and escolar. In recent 
years escolar has become a commercial species and is 
now reported in the logbook, but this was not the case 
in the early part of the time period examined. Escolar 
is sometimes locally called oilfish or walu, but oilfish is 
actually the common name for Ruvettus pretiosus which 
represents a relatively rare species in the catch of the 
longline fishery. 
For the 10 species fully reported in the logbooks, we 
computed a monthly catch-per-unit-of-effort (CPUE) 
time series. Logbook monthly CPUEs were computed as 
the total number of fish of a species caught in a month 
divided by the total number of hooks multiplied by 
1000; thus CPUE was computed as the number of fish 
per 1000 hooks. A generalized additive model (GAM) 
(Hastie and Tibshirani, 1990) was then used over the 
1996-2006 period that fitted observed monthly CPUE 
and contained a linear function of year to model the 
time trend, a smoothed monthly term to model the sea- 
sonal pattern, and a smoothed spatial term computed 
from mean monthly latitude and longitude to incorpo- 
rate any spatial contribution to CPUE. 
For the three species not fully reported in the log- 
books (lancetfish, snake mackerel, and escolar), we used 
observer catch and effort data which covered about 16% 
of the fishing effort over the decade. Because we had 
much less observer coverage than logbook coverage, we 
pooled the observer data over the year and computed 
an annual, rather than monthly, CPUE time series. 
Observer annual CPUEs were computed as the total 
number of fish of a species caught in a year on vessels 
carrying observers divided by the total number of hooks 
used by those vessels multiplied by 1000. Because of the 
limited data points with our annual CPUE time series, 
a simple linear regression was fitted to the annual 
CPUE data. Although the limited observer coverage was 
considerably less than that reflected by the logbook data 
