142 
Fishery Bulletin 120(2) 
Genetic analyses 
Genetic identification of opah was completed as described 
in Hyde et al. (2014). Briefly, Hyde et al. (2014) used sam- 
ples of fin tissue collected either by fishery observers at sea 
during research cruises or by researchers after whole fish 
were landed in port (in either Hawaii or California) and pre- 
served in 95% ethanol until processed. From each sample, 
DNA was extracted by using a boiling protocol with Chelex 
100? resin (Bio-Rad Laboratories Inc., Hercules, CA), and a 
portion of the mitochondrial cytochrome c oxidase gene was 
amplified, sequenced, and compared to reference sequences 
for bigeye Pacific opah (accession no. JF931871, GenBank, 
available from website) and smalleye Pacific opah (acces- 
sion no. JF931880, GenBank, available from website) to 
identify species. 
This study incorporated samples collected by research- 
ers and observers and used in the Hyde et al. (2014) study 
and samples collected by commercial fishermen since that 
paper was published. For samples collected by observ- 
ers, sampling locations were identified to specific sets, 
whereas the more recent samples collected by fishermen 
were associated only with specific trips, not with individ- 
ual sets. To standardize the spatial resolution of sampling 
of tissue for genetic analysis, location data for all samples 
were aggregated to trip level by taking the average lati- 
tude and longitude of all sets within the trip in which opah 
were landed (an average of 13 sets per trip), on the basis 
of logbook data from the same sampled trip. There were 
141 deep-set and 2 shallow-set longline trips associated 
with genetic samples of opah. Given the lack of genetic 
samples from shallow-set trips, only the genetic samples 
from deep-set trips were retained for further analysis. In 
addition, 2 deep-set trips had a difference of more than 
5° between the maximum and minimum latitude or longi- 
tude of a trip, and they were excluded, yielding a total of 
139 deep-set longline trips that were used for estimating 
distribution of opah species. 
Spatial distributions of species 
Based on the genetic analyses described above, the pro- 
portion of bigeye Pacific opah versus smalleye Pacific 
opah within 5°-by-5° grid cells were calculated. A series 
of GAMs were then developed by using the mcgv pack- 
age (vers. 1.8-33; Wood, 2017) in R, vers. 3.6.1 (R Core 
Team, 2019), to explain the differences in the proportions 
of bigeye Pacific opah. Latitude (Jat) and longitude (lon) 
were used to predict the proportion of bigeye Pacific opah 
(propnBigeye) through beta regression, weighted by the 
total number of opah caught (sample size) within each 
grid cell: 
propnBigeye ~ s(lon, lat), 
where s = a thin-plate regression spline (Wood, 2017). 
* Mention of trade names or commercial companies is for identi- 
fication purposes only and does not imply endorsement by the 
National Marine Fisheries Service, NOAA. 
We tested the inclusion of season (quarter of the year) as 
an additional predictor in the GAMs, to assess whether 
spatial patterns in species distributions varied through- 
out the year. However, the GAMs that include season were 
not superior to those without that variable, with <1% dif- 
ference in variance explained and higher Akaike informa- 
tion criteria. As such, for further analysis, we proceeded 
with GAMs that include just latitude and longitude as 
predictors. 
Three GAMs were built by using identical predictor and 
response variables but different knot () values for the iso- 
tropic thin-plate regression spline: 4 (low), 6 (medium), and 
8 (high). We chose to vary k to capture some of the uncer- 
tainty in the spatial prediction surface because the loca- 
tions of genetically identified opah did not cover the full 
geographic range of the pelagic longline fleet. Values of 4, 
6, and 8 allowed different patterns of spatial extrapolation 
outside the sampled area, without obvious overfitting. 
The fitted GAMs were then used to generate maps with 
the expected proportions of bigeye Pacific opah over the 
study area and were used to identify areas dominated 
by either bigeye Pacific opah or smalleye Pacific opah, by 
using an arbitrary threshold of 0.7. Areas with expected 
proportions of bigeye Pacific opah >0.7 and <0.3 were clas- 
sified as dominated by bigeye Pacific opah and smalleye 
Pacific opah, respectively. Areas with expected proportions 
between 0.3 and 0.7 were considered mixed areas with- 
out dominance of a single species of opah. The logbook 
data from the deep-set longline fishery were divided into 
3 subsets—data for areas dominated by bigeye Pacific 
opah, data for areas dominated by smalleye Pacific opah, 
and data for mixed areas—and were used to calculate the 
number of sets and average CPUE in each type of area. 
To estimate the uncertainty of the GAM predictions, 
an iterative jackknife procedure was performed. Jack- 
knifing resamples data with a leave-one-out method; the 
medium (k=6) GAM with 139 trip-level data points was 
rerun 139 times, each time with a single trip removed 
(with replacement) to predict species proportion for each 
1°-by-1° block in a grid spanning the study area. To visu- 
alize the uncertainty of the GAM predictions, the mean, 
standard deviation, and coefficient of variability were 
estimated for each block, with the coefficient of variabil- 
ity calculated as (standard deviation/mean)100. 
Results 
Spatial and temporal patterns: fishing effort and catch 
per unit of effort 
A compilation of all data across the time series reveals 
differences in the deep- and shallow-set longline fisheries. 
Effort in the deep-set fishery increased from roughly 375 
sets in 1996 to 18,700 sets in 2018. In contrast, effort in 
the shallow-set fishery peaked in 1998 at roughly 4700 sets. 
The shallow-set fishery was temporarily closed from 2001 
through 2004 (Federal Register, 2004), and annual effort 
remained below 2000 sets per year from 2005 through 2018. 
