Misa et al. : Establishing species-habitat associations for 4 eteiine snappers 
297 
dance estimate that avoids the potential problem of 
counting the same fish multiple times as it re-enters a 
camera’s field of view. Many studies have determined 
that MaxNo is positively correlated with fish density 
(Ellis and DeMartini, 1995; Priede and Merrett, 1996; 
Willis et ah, 2000; Willis and Babcock, 2000; Yau et 
al., 2001; Cappo et al., 2003). This parameter also has 
been found to be highly correlated with the traditional 
parameter of CPUE used in fishing surveys (Ellis and 
DeMartini, 1995). MaxNo was recorded for all fishes 
present in the BotCam video footage, but only data for 
the 4 species of interest were analyzed. 
Permutational analysis of variance (PERMANOVA) of 
the data was performed in Primer 6 (PRIMER-E Ltd., 
Ivybridge, UK) with PERMANOVA+ (Anderson et al., 
2008). With PERMANOVA, the data are not assumed 
to be normally distributed; therefore, this technique was 
deemed appropriate for analysis of our data, which in- 
cluded a highly skewed (overdispersed) relative abun- 
dance distribution due to an unbalanced experimental 
design and frequent zero counts. The 4 species consid- 
ered in our study do not all occupy the entire depth range 
sampled (Polovina et al., 1985; Haight, 1989; Everson 
et al., 1989; Merritt et al., 2011). To constrain the data 
to an appropriate range for each species, the depths at 
which each species had its greatest MaxNo had to be 
identified. For the initial analysis, depth was divided 
into 30-m bins from 90 to 300 m. Relative abundance 
values were square-root transformed to compensate for 
numerous zero counts and occasional large numbers. 
A Euclidean distance matrix was used in the statisti- 
cal test with a type-III sum of squares. If a significant 
difference (P<0.05) was observed across depth bins, a 
subsequent pair-wise PERMANOVA was performed to 
determine the preferred depths of each species. Subse- 
quent analyses (MaxNo and fork length [FL] ) were then 
constrained to the depth preferences identified for each 
of the 4 species studied. 
Through identification of habitat preferences, the 
influence of BRFA location (i.e., combined area inside 
and outside a BRFA) and protection (i.e., area inside 
versus outside a BRFA) could not be overlooked. PER- 
MANOVA in a 3-way crossed design was used to deter- 
mine how BRFA location (BR, 6 levels, fixed), protec- 
tion (PR, 2 levels, fixed), habitat type (HA, 4 levels, 
fixed), and the interaction of these factors affected the 
relative distribution of each species. MaxNo values 
were square-root transformed, and the PERMANOVA 
was run on a Euclidean distance matrix with type-III 
sum of squares. Where significant results (P<0.05) oc- 
curred, pair-wise testing was performed to identify spe- 
cific differences. 
For individual fish visible in both BotCam cam- 
eras, FL was measured with stereo-photogrammetric 
measurement software: Visual Measurement System 
7.5 (Geometric Software Pty. Ltd., Coburg, Victoria, 
Depth (m) 
Figure 2 
Mean relative abundance (MaxNo) with standard error (SE) across 7 depth bins for Opakapaka ( Pristi - 
pomoides filamentosus), Kalekale (P. sieboldii), Onaga (Etelis coruscans), and Ehu ( E . carbunculus) from 
surveys of these species conducted in the main Hawaiian Islands from May 2007 to June 2009 with the use 
of a baited stereo-video camera system. Columns with the same letter are not significantly different from 
each other (P>0.05, post hoc permutational analysis of variance [PERMANOVA] testing). Error bars indicate 
±1 SE of the mean. 
