Merritt et al. : BotCam: a baited camera system for nonextractive monitoring of bottomfish species 
65 
hard substrate; therefore in any sampling there will 
be an aggregated distribution rather than a random or 
uniform one (Haight et ah, 1993a; Kelley and Ikehara, 
2006). Indeed, the present results show that MaxNo, 
similar to many other types of count data, were not 
normally distributed; many camera deployments re- 
sulted in zero fish and others with up to 29 fish (Fig. 4). 
MaxNo appears to be a more appropriate metric than 
TFA for estimating relative abundance in this case, but 
will likely require analysis with statistical models that 
are designed for nonuniform dispersion patterns. 
Knowledge of the distribution of fishes among habi- 
tats is of importance to fisheries management, and 
such information can readily be obtained with the Bot- 
Cam system. The distributions of E. coruscans and P. 
filamentosus among depth bins and habitat substrata 
types in our study (Fig. 5) indicate that E. coruscans 
on Penguin Bank prefer high slopes and deeper water, 
whereas P. filamentosus do not have a strong prefer- 
ence for a particular bottom type but are found in the 
shallowest three quarters of the depth range sampled. 
Modeling the distribution of both species across depth, 
slope, and substrate type indicated that these factors 
were important in understanding the association of 
these species with their habitat. Currently, the es- 
sential fish habitat for these species is simply defined 
as all waters between 100 and 400 m deep. Although 
beyond the scope of this study, the results show that 
additional work with BotCam would enable fisher- 
ies scientists to more accurately define essential fish 
habitats and habitat areas of particular concern on a 
species-by-species basis. Combined with direct observa- 
tion of habitat, BotCam is also a tool that will allow 
for a much finer resolution of habitat classification (i.e., 
bedrock versus boulders versus cobbles) and enable 
species preferences to be discerned (see Stoner et al., 
2008). Parrish et al. (1997) applied this technique to 
investigate habitat affinity of juvenile P. filamentosus 
and identified premium habitat by using direct observa- 
tions from video cameras. 
One objective of this study was to evaluate the preci- 
sion and accuracy of the stereo-photogrammetric tech- 
nique for obtaining accurate size measurements of 
bottomfishes. After analyzing repeated measurements 
of E. coruscans and P. filamentosus, a discrepancy was 
apparent between the species. The smaller number of 
E. coruscans measured and the larger standard de- 
viation of the measurements relative to P. filamento- 
sus were likely the result of E. coruscans being found 
in deeper water, where visibility and image quality 
decrease, making video measurement more difficult. 
Nonetheless, valuable information about the size distri- 
bution of these fishes was collected (Fig. 6), indicating 
that BotCam could be useful as a nonextractive tool 
for sampling size distributions for stock assessment. 
Additional experience in both calibrating the camera 
system and in using the stereo-video software will 
improve the precision and accuracy of size measure- 
ments as evidenced by previous studies where a similar 
system and software were used (Harvey et al., 2003). 
Harvey et al. (2002) compared fish length estimates 
from stereo-video and scuba divers and found video to 
provide consistently more accurate and precise data. 
Additionally, Harvey et al. (2010) conducted a similar 
study on the accuracy and precision of stereo video 
camera system and found that the length of the object 
measured was a major factor in reducing variance dur- 
ing measuring. In contrast to this finding, we suggest 
that size was not a factor, although our study supports 
the finding that precision degrades with distance away 
from the camera. 
The size distributions of P. filamentosus and E. cor- 
uscans estimated in our study were consistent with 
published data for both species. Haight et al. (1993a) 
estimated the length at maturity of P. filamentosus to 
be 430 mm, and maximum length to be 780 mm. Our 
estimates for P. filamentosus ranged from 344 mm to 
660 mm, normally distributed throughout the reported 
size range (Fig. 6). Everson et al. (1989) estimated the 
length at maturity of E. coruscans to be 663 mm, and 
maximum length to be 925 mm. Our estimates for E. 
coruscans ranged from 432 mm to 832 mm, again nor- 
mally distributed across the reported size range (Fig. 
6). These results indicate that BotCam can estimate 
relative size frequencies, both pre- and post-sexual 
maturity and therefore could be used for monitoring 
recruitment and changes in spawning potential ratios. 
In neither species was a fish measured near its re- 
ported maximum size. The reasons for this could be low 
sampling effort, size-related differences in behavior or 
habitat use, bias caused by measuring only at MaxNo, 
or simply that individuals of such large size were absent 
from the sampled area. Juveniles of these species were 
also absent from the video recordings, possibly because 
they remained close to the bottom near cavities because 
of their vulnerability to predation, as typical of other 
bottom associated fishes. Juveniles could have been in 
the vicinity of BotCam, but because of the presence of 
larger fish, such as S. dumerili, were possibly unwilling 
to come up to the cameras. 
Monitoring deepwater fishes and their habitat is a 
difficult and costly undertaking. We tested the effec- 
tiveness of a new baited stereo-video camera system 
(BotCam) and found it an efficient tool in places where 
diver surveys are impossible and ROV or submersible 
surveys are cost prohibitive or provide data of uncer- 
tain quality (Kelley et al., 2006; Stoner et al., 2008). 
The success rate of data collected per deployment in 
this study supports the use of BotCam for studying 
biologic assemblages at depths ranging from 0 to 300 
meters. As a nonextractive method, BotCam could prove 
particularly valuable in marine protected areas, where 
restrictions on fish removal may limit the usefulness of 
traditional sampling methods (Willis et al., 2003; Denny 
et al., 2004; Willis and Millar, 2005). Future work must 
include careful calibration of BotCam data with tradi- 
tional population assessment data, including measures 
of relative abundance based on fisheries-dependent data 
such as CPUE. In addition, calibration with other non- 
extractive methods, such as acoustic surveys, is needed. 
