296 
Fishery Bulletin 111(4) 
data from these cruises have not been made publicly 
available, with the exception of the bathymetry data 
that have been incorporated into a 50-m-resolution syn- 
thesis of the entire MHI that is available from the Ha- 
waii Mapping Research Group (http://www.soest.hawaii. 
edu/hmrg/multibeam/index.php). 
Multibeam backscatter data in grids with a 20-m 
resolution cannot be used effectively to identify specific 
substrate types, such as mud, sand, pebbles, cobbles, 
boulders, and bedrock, because more than one of these 
substrate types often can be found on the seafloor in 
an area of 20x20 m. Similarly, more than one type of 
slope can be found in areas of that size because of the 
presence of small carbonate ledges, large boulders and 
blocks, sand dunes, and other small-scale topographic 
features common to seafloors in the Hawaiian Archipel- 
ago. Multibeam data values for each grid cell (20x20 m) 
are typically derived through calculation of either the 
Gaussian weighted means (bathymetry) or the medians 
(backscatter) of the sonar footprints within each cell. 
For these reasons, only 4 general habitat types were de- 
rived from these multibeam data: hard substrate with 
high slope (hard-high), hard substrate with low slope 
(hard-low), soft substrate with high slope (soft-high), 
and soft substrate with low slope (soft-low). Bathymetry 
data from the different sonar systems generally were 
consistent. 
After a number of slope analyses were conducted in 
ArcGIS 9.1 (Esri, Redlands, CA), a value of 20° was de- 
termined to be a reasonable boundary between the high 
and low slopes that appeared in the bathymetry images. 
Backscatter data, however, are often inconsistent be- 
tween systems with different frequencies. Furthermore, 
the backscatter data used in this study were processed 
in different ways by different technicians. As a result, 
boundary values between hard and soft substrates had 
to be determined on a basis of per system and per cruise. 
A value of 187 was used as the boundary between hard 
and soft substrates for the EM 300 data and was vali- 
dated through examination of video from submersible 
surveys. Boundary values for the EM 1002 data ranged 
from -41 to 150 and were established through compari- 
son of areas of overlap with EM 300 data and analysis 
of video from submersible surveys. 
Habitat was classified at a resolution of 200x200 m 
for areas in and around BRFAs. Polygons for high and 
low slopes and hard and soft substrates were generated 
with the Raster calculator in ArcGIS 9.1. Intersects of 
slope and hardness resulted in polygons for the 4 hab- 
itat types. A grid cell (200x200 m) was superimposed 
over these polygons, and the areas of the habitat types 
within each grid cell were calculated. Each grid cell was 
assigned a habitat type on the basis of which habitat 
type was observed in the greatest proportion in that 
area. 
A stratified-random sampling approach was used to 
select locations for BotCam sampling. Although the pur- 
pose of our study was to evaluate species-habitat as- 
sociations, another goal of this project was to evaluate 
population changes inside and outside of BRFAs. This 
objective affected our sampling design. We used data 
from 625 deployments of the BotCam conducted inside 
and outside of 6 of the 12 current BRFAs (Fig. 1) be- 
tween May 2007 and June 2009. The 6 BRFAs that were 
sampled are located off Niihau (BRFA B), Kaena (BRFA 
D), Makapuu (BRFA E), and Penguin Bank (BRFA F), in 
Pailolo Channel (BRFA H), and outside of Hilo (BRFA 
L). The Niihau and Hilo BRFAs were areas of contin- 
ued closure from the initial implementation of BRFAs 
in 1998. The Makapuu and Penguin Bank BRFAs were 
expanded versions of smaller preexisting BRFAs from 
1998, and the BRFAs off Kaena and in Pailolo Channel 
were areas newly closed in 2007. 
The BotCam was lowered to depths of 100-300 m. 
Although the EFH for deep bottomfishes in Hawaii ex- 
tends to 400 m, the video cameras work under ambient 
light to only 300 m, thus limiting the depth range of our 
sampling. Sampling effort was weighted toward known 
preferred bottomfish habitats to ensure greater replica- 
tion where fish densities were expected to be higher. 
Because previous studies have found bottomfishes asso- 
ciated with hard substrates, high slopes, or a combina- 
tion of both (Polovina et al., 1985; Ralston et al., 1986; 
Haight et al., 1993a; Parke, 2007), for our study, hard- 
high habitats were considered the most suitable and 
soft-low habitats the least suitable. To sample a BRFA, 
32 BotCam deployments inside and 32 outside but ad- 
jacent to a BRFA were completed over grids of each 
habitat type with the following replication: 12 hard- 
high, 8 hard-low, 8 soft-high, and 4 soft-low. BotCam 
deployments targeted centroids of randomly selected 
grid cells (200x200 m) and were kept a minimum of 400 
m apart to reduce the likelihood of sampling overlap. 
In regions where a given habitat type was not pres- 
ent, sampling intensity was increased in the next most 
suitable habitat. This approach led to skewed sampling 
across habitat types in Pailolo Channel because only 
low-slope habitats were identified at a resolution of 
200x200 m. When BotCam deployments did not yield 
usable video (e.g., no recordings or extremely dark im- 
agery), the BotCam was redeployed at that location on 
another day. As often happens during sampling efforts 
in the field, not all targeted grids were sampled because 
of weather and equipment issues. In the 2-year sam- 
pling period covered by this study (2007-09), 4 of the 
6 BRFAs (Niihau, Makapuu, Penguin Bank, and Pailolo 
Channel) were sampled twice and the Kaena and Hilo 
BRFAs were sampled only once. 
BotCam video footage was reviewed in the labora- 
tory to estimate the relative abundance, recorded as the 
maximum number of a particular species observed in a 
single frame of video (MaxNo), of Opakapaka, Kalekale, 
Onaga, and Ehu with VF Deep Portal (Deep Develop- 
ment Corp., Sumas, WA) and Adobe Premiere Pro CS4 
(Adobe Systems, Inc., San Jose, CA) software programs. 
Fishes were identified to the most specific taxonomic 
classification possible with a species identification ref- 
erence (Randall, 2007). MaxNo is a conservative abun- 
