366 
Fishery Bulletin 106(4) 
attached to ensure the gear is fished on the bottom. Each 
station consists of 160 skates set across depth contours 
from 150 to 1000 m. At each station there are 7200 hooks 
set over 16 km. The objective of the setting pattern is to 
evenly distribute sampling effort over the depths that 
sablefish inhabit. Gear retrieval begins after the gear 
has soaked for 3 hours, and fish and baited hooks are 
counted as they are brought aboard. Catch rates are tab- 
ulated as the number of fish of each species per skate of 
gear (no. fish/45 hooks). Bottom depth is recorded every 
five skates as the gear is hauled. Interpolated depths are 
assigned to skates that do not have a recorded depth. 
Shortraker and rougheye rockfish catches were pooled 
for this analysis because they are not distinguished 
from one another during gear retrieval. These species 
are very similar in appearance and difficult to distin- 
guish, share the same habitat on the upper continental 
slope, and often are found in fishery hauls (Clausen 
and Fujioka, 2005). In this article we will refer to both 
species as rockfish. 
AFSC trawl surveys The AFSC groundfish trawl sur- 
veys sample the continental shelf and upper continental 
slope of the Gulf of Alaska at depths to 500 m, and 
in some years as deep as 1000 m, during the summer 
(Britt and Martin, 2001; Fig. 1). The surveys follow a 
stratified random sampling pattern with 49 strata that 
are categorized by depth, geological area (e.g., gully, 
slope), and management area (i.e., latitude and longi- 
tude). Trawls are hauled at a vessel speed of 3 knots 
for 15 minutes, or in some years, for 30 minutes, and 
tows are conducted along a constant bottom depth. An 
average depth is assigned to each trawl haul. All fish 
caught on the surveys are enumerated and a catch rate 
is calculated by dividing the number of fish caught by 
the area swept by the net (no. fish/km 2 ). 
Analysis 
Longline correlations Average catch rates from the 
1979-2003 annual longline surveys were calculated by 
50-m depth increments to determine the preferred depth 
range for grenadier and rockfish in each sablefish man- 
agement area. Analyses were done separately by man- 
agement area because preferred depth ranges and catch 
rates have differed in each area for each fish species. The 
preferred depth ranges of grenadier and rockfish in each 
area were defined as the range where the average catch 
rate, in all years, was at least 20% of the highest catch 
rate. This method resulted in disregarding depths where 
the catch rate was less that 20% of the average peak 
catch rate. Our intention was to consider only depths 
where grenadier or rockfish were prevalent. An average 
catch rate was then calculated for the preferred depth 
range for grenadier and rockfish in each management 
area in each year at each longline survey station. Catch 
rates were also calculated for sablefish in the grenadier 
and rockfish preferred depth ranges so that sablefish 
catch rates could be compared to grenadier and rockfish 
catch rates. Preferred depth ranges were calculated from 
longline survey catch rates instead of trawl survey catch 
rates because there is little fishing effort below 500 m on 
the trawl surveys. In addition to calculating the average 
catch rate for the whole preferred depth range, we also 
calculated the average catch rate by 50-m increments 
within the preferred depth range in order to make more 
fine-scale species comparisons. 
Before tests for statistical significance, catch rates 
were transformed by using natural logarithm, square 
root, or fourth root transformations to help meet as- 
sumptions of normality. Pearson’s correlation coefficients 
and tests of significance were calculated. Catch rates 
calculated as catch per skate did not meet the assump- 
tions of normality due to a prevalence of zeros. There- 
fore, average catch rates of the preferred depth range 
by station were used to calculate correlation coefficients. 
Longline versus trawl correlations Because longline 
and trawl data are collected at the same time, areas, 
depths, and from the same habitats, trawl correlations 
are a good comparison for longline correlations. In the 
trawl data, a positive or zero correlation indicates that 
species do not have differing habitat preferences and 
that the species do not directly compete with each other. 
A negative correlation on longline gear and a positive or 
zero correlation in trawl gear would indicate that the 
negative correlations in longline gear could be caused by 
hook competition and not by these other factors. 
Trawl catch rates were computed for each haul for 
each species in each year the trawl surveys occurred 
(1984, 1987, 1990, 1993, 1996, 1999, 2001, 2003, and 
2005). Because depth ranges varied throughout the 
Gulf of Alaska, the entire depth distribution of grena- 
dier and rockfish throughout the Gulf of Alaska was 
included in this analysis. The depth distribution was 
rounded to the nearest 50 m because longline survey 
catches were already summarized by 50 m increments. 
For grenadier the range was 300-700 m and for rock- 
fish the depth range was 250-550 m. Catches deeper 
than 700 m were not included because of the limited 
trawl survey effort below this depth. Trawl catch rates 
were also transformed to help meet the assumptions of 
normality. In one case, namely for the rockfish trawl 
data, the data could not be transformed to fit a normal 
distribution, but the transformation was retained be- 
cause it aided in the visual analysis of the data. In this 
instance, nonparametric correlation tests, Spearman’s 
p and Kendall’s r, were used to calculate correlation 
coefficients and to test for significance. Longline catch 
rates were summarized to be comparable with trawl 
catch rates; average catch rates were computed for the 
same depth ranges and years at each station. 
Results 
Longline catch rates by area 
Preferred depths of grenadier and rockfish differed by 
management area (Table 1). Peak catch rates also dif- 
