298 
Fishery Bulletin 1 1 1 (4) 
Table 1 
Results of permutational analysis of variance (PERMANOVA) with relative abundance (MaxNo) data from our sur- 
veys of 4 species — Opakapaka t Pristipomoides filamentosus), Kalekale (P. sieboldii), Onaga ( Etelis coruscans) , and 
Ehu (E. carbunculus) — in the main Hawaiian Islands between May 2007 and June 2009. The following factors were 
tested within the preferred depths of each species: bottomfish restricted fishing area location (BR), protection (PR), 
and habitat type (HA). Preferred depths are noted in the column head for each species. df=degrees of freedom; 
P=PERMANOVA P-statistic; P=PERMANOVA P-value. Asterisks indicate statistical significance at P<0.05. 
Factor 
Opakapaka 
(90-210 m) 
Kalekale 
(180-270 m) 
Onaga 
(210-300 m) 
Ehu 
(210-300 m) 
df 
F 
P 
df 
F 
P 
df 
F 
P 
df 
F 
P 
BR 
5 
2.86 
0.02* 
5 
2.07 
0.09 
5 
1.54 
0.17 
5 
4.78 
0.00* 
PR 
1 
0.00 
1.00 
1 
0.07 
0.79 
1 
0.07 
0.78 
1 
0.31 
0.58 
HA 
3 
8.28 
0.00* 
3 
1.68 
0.18 
3 
3.87 
0.02* 
3 
2.83 
0.04* 
BRxPR 
5 
0.63 
0.66 
5 
0.55 
0.72 
5 
0.56 
0.70 
5 
0.81 
0.54 
BRxHA 
13 
0.64 
0.80 
12 
1.89 
0.06 
13 
0.69 
0.71 
13 
2.33 
0.01* 
PRxHA 
3 
0.62 
0.59 
3 
0.87 
0.45 
3 
0.56 
0.62 
3 
0.93 
0.42 
BRxPRxHA 
12 
1.02 
0.42 
10 
0.44 
0.91 
9 
0.59 
0.76 
9 
0.58 
0.79 
Residual 
247 
282 
295 
295 
Australia) and PhotoMeasure 1.74 (SeaGIS Pty. Ltd., 
Bacchus Marsh, Victoria, Australia). Measurements of 
individual fish were taken at the point of MaxNo or 
at the point in the video where the most fish could be 
measured to ensure that individuals were not repeat- 
edly measured at various times during video analy- 
sis. Replicate measurements were taken for individual 
fish to increase the accuracy of the measurement. An 
LED device was used to ensure synchronicity of the 
video footage from the left and right cameras. A root- 
mean-square error or residual parallax >10 mm and 
a precision-to-FL ratio >10% were indicative of inac- 
curate measurements. To ensure the quality of fish 
length data, these measurements were removed from 
the analyses in this study. The same 3-way crossed 
design from the PERMANOVA of relative abundance 
(BR, PR, HA) was used to test FLs for each species. 
Transformation of FLs, however, was not necessary be- 
cause these data typically were normally distributed. 
Because only variations in mean length were evalu- 
ated with the previously described approach, additional 
analyses were undertaken to investigate size-related 
changes in habitat association. A linear regression was 
used to evaluate the relationship between depth and 
FL for each of the 4 species studied to identify ontoge- 
netic shifts with depth. As part of our examination of 
ontogenetic shifts across habitat types, a contingency 
table (tested with Pearson’s chi-square test) was used 
to determine whether the size-class distribution of each 
species was independent of habitat type. Fork lengths 
were grouped into 10-cm bins. This size interval was 
chosen to maximize the number of observations in each 
size bin. Merritt et al. (2011) tested and found mea- 
surements from BotCam video to be accurate to within 
0.3-0. 9 cm, making such a grouping very robust. 
Results 
For all 4 species studied, significant differences in rel- 
ative abundance were found across depth bins (PER- 
MANOVA, P<0.05). Pair-wise comparisons of MaxNo 
from the 7 depth bins highlighted the depth preference 
of each species (Fig. 2). MaxNo was highest from 90 to 
210 m for Opakapaka (post hoc PERMANOVA, P<0.05). 
The preferred depths of Kalekale were 180-270 m, and 
both Onaga and Ehu had the deepest range among spe- 
cies at 210-300 m (post hoc PERMANOVA, P<0.05). 
Within the preferred depths of a species, either 
BRFA location, habitat type, or the interaction of these 
2 factors had an effect on the relative abundance of 3 
of the 4 species studied (Table 1). Protection and the 
interaction of all other factors with protection, how- 
ever, did not have an effect (PERMANOVA, P>0.05). 
BRFA location and habitat type were each significant 
factors for Opakapaka. Hilo had the highest relative 
abundance of this species among sampled locations, 
and hard-low habitats yielded greater abundance esti- 
mates for Opakapaka than other habitat types (Fig. 3; 
post hoc PERMANOVA, P<0.05). Although no signifi- 
cant location or habitat effects were observed for Kale- 
kale, the interaction of BRFA location and habitat type 
was marginal (P=0.06; Table 1); 2 of the largest counts 
of this species (100 and 85 individuals) occurred on 
hard-high habitats at Niihau and led to a high mean 
MaxNo (Fig. 3). 
Habitat type was the only factor that affected the 
relative abundance of Onaga. Hard substrate habitats, 
with either high or low slope, had greater mean MaxNo 
for Onaga than soft substrate habitats (Fig. 3; post hoc 
PERMANOVA, P<0.05). BRFA location, habitat type, 
and the interaction of these 2 factors were significant 
