64 
Fishery Bulletin 109(1 ) 
that an area-based bait plume model worked well to 
explain variation in their count data but were unable to 
determine if the correlation between counts and current 
was a result of the bait plume size or an indication of 
the preferred habitat of the fishes. Further work with 
BotCam is necessary to evaluate the area of influence 
of the bait, but the skewed relationship between MaxNo 
and TFA (Fig. 4) indicates that attraction to the bait is 
rapid and, therefore, local in its effect. 
Another confounding factor is the visual attraction 
of fish to the camera system itself. Watson (2005) refer 
to this as the “curiosity” effect and although it is a dif- 
ficult value to quantify, it is clear from the video record- 
ings that fish do react to the camera system. Unbaited 
deployments need to be carried out to better understand 
the magnitude of this effect. 
Baited camera systems have historically been used 
to determine either TFA or MaxNo to estimate relative 
density of the attracted fishes (Bailey et al., 2007). In 
— 
many studies, TFA has been used in an inverse-square 
model as a metric of abundance (Priede et al., 1994). 
It is assumed with the use of TFA that individuals 
are uniformly distributed in space, act independently 
of each other (i.e., there is no schooling behavior), all 
fishes that contact the odor plume swim up current to 
the camera, and the effect of the bait plume on fish 
counts is linear and dependent on local current speed. 
Thus, short TFAs imply greater densities than long 
TFAs. In more recent statistical models, the arrival 
rate instead of the TFA has been used, which allows 
an estimate of a confidence interval (Farnsworth et 
ah, 2007), but both measures are based on the same 
basic assumptions. These metrics have been applied 
primarily to deep sea fishes (>1000 m) inhabiting low- 
energy, bathymetrically monotonous environments 
(Priede and Merrett, 1996). They are also hypersensi- 
tive at rapid TFAs (<~5 min) and insensitive at long 
TFAs (> —120 min; King et al., 2006; Yeh and Drazen, 
2009). Shallower water environments, such 
as those surveyed in the current study, are 
more dynamic ecologically and physically 
than in the deep sea and therefore fishes 
tend to be less evenly distributed in space. 
The assumptions about the uniform dis- 
tribution of the target fishes or linearity 
of responses to the odor plume required by 
TFA models often cannot be met. As a result, 
studies examining shallow-water fishes (El- 
lis and DeMartini, 1995; Willis et al., 2000; 
Watson et al., 2005; Kelley and Ikehara, 
2006; Stoner et al., 2008) have used MaxNo 
as an index of relative density which avoids 
the potential for recounts of the same fish as 
they exit and reenter the field of view dur- 
ing the survey period. Ellis and DeMartini 
(1995) found that MaxNo is positively corre- 
lated to catch per unit of effort (CPUE) and 
concluded that it is a useful index of abun- 
dance. Likewise, Stoner et al. (2008) con- 
cluded that MaxNo was the optimal measure 
because it is correlated with seine hauls and 
is consistent across habitat types. Willis et 
al. (2000) compared a baited camera system 
with visual surveys and angling surveys and 
also concluded that video survey techniques 
with MaxNo provided reliable estimates of 
relative density. In the present study, TFAs 
were very short (Fig. 4) and could produce 
highly variable and spuriously high esti- 
mates of abundance (King et al., 2006). This 
is associated with the lack of sensitivity of 
TFA to small densities where arrival time 
is dependent on the position and response 
to bait of the closest fish. We assumed that 
the bait plume was not uniform because of 
the variability in conditions (i.e., currents) 
and rugged bathymetry. Furthermore, it is 
well known that some species of bottomfish 
school, whereas others associate only with 
Fork length (mm) 
Figure 6 
Length-frequency distribution of (A) Etelis coruscans and (B) Pris- 
tipomoides filamentosus from BotCam deployments at Penguin 
Banks, Hawai’i, between June 2006 and February 2007 as measured 
by stereo-video software Vision Measurement System (Geomsoft, 
Victoria, Australia). Only fish identified at the time of MaxNo 
(maximum number of individuals in a single frame) were measured. 
Each fish seen around the time of MaxNo was measured six times 
(from six different frames of the video) in order to tease out errors 
due to fish motions and human error. The average fork lengths are 
binned in 50-mm intervals. 
