344 
Fishery Bulletin 115(3) 
techniques have been applied to estimate the size 
selectivity of a survey trawl (Somerton et al., 2007) 
and even of the absolute catchability of a trawl sur- 
vey (Somerton et al., 2013). Estimating the sampling 
efficiency of a survey method or the catchability of a 
survey can be important for stock assessments because 
estimates of these quantities can be incorporated into 
stock assessment models (Maunder and Punt, 2013) as 
Bayesian priors to improve model reliability. Inclusion 
of such priors in stock assessment models is especially 
important for developing surveys, like those envisioned 
for camera vehicles, because the survey time series will 
initially be too short to adequately inform the model 
(Somerton et al., 1999). 
To help develop methods for estimating the sam- 
pling efficacy of camera vehicles and to understand 
better the stimuli produced by these vehicles that may 
influence sampling efficiency, the U.S. National Marine 
Fisheries Service initiated the Untrawlable Habitat 
Strategic Initiative, which conducts experiments to ob- 
serve and quantify the reactions of fish to remotely op- 
erated vehicles, autonomous underwater vehicles, and 
towed camera vehicles. For the first experiment, which 
is considered here, an observational test bed was de- 
veloped that consisted of unlit, bottom-mounted stereo 
cameras set in shallow, clear water to unobtrusively ob- 
serve the responses of fish to various types of camera 
vehicles. The study site was a sponge and soft coral 
reef that is situated in the northeastern Gulf of Mexico 
and is inhabited by a variety of tropical fishes, includ- 
ing several species of snappers and groupers important 
to commercial and recreational fisheries (Coleman et 
al. 1 ). 
Several independent strategies were used to quan- 
tify the responses of fish. One of these was to mea- 
sure changes in the relative density of fish in the area 
viewed by single cameras (Campbell 2 ) and by DID- 
SON 3 imaging sonars (Sound Metrics, Bellevue, WA) 
(Wakefield 4 ) as the benthic cameras were approached 
by camera vehicles. Another strategy, described here, 
was to quantitatively assess fish behavior in response 
to the camera vehicles by using stereo images to re- 
peatedly measure the 3-dimensional positions of indi- 
vidual fish over time (i.e., target tracking). This type of 
information, in turn, allowed measurement of changes 
in swimming speed and direction of individual fish as 
1 Coleman, F., G. Dennis, W. Jaap, G. P. Schmahl, C. Koenig, 
S. Reed, and C. Beaver. 2004. Part I: Status and trends 
in habitat characterization of the Florida Middle Grounds. 
Final report to the National Oceanic and Atmospheric Ad- 
ministration Coral Reef Conservation Grant Program, 135 
p. [Available from website. 
2 Campbell, M. D, 2016. Unpubl. data. Mississippi Labora- 
tories, Southeast Fish. Sci. Cent., Natl. Mar. Fish. Serv., 3209 
Frederic St. Pascagoula, MS 39567. 
3 Mention of trade names or commercial companies is for iden- 
tification purposes only and does not imply endorsement by 
the National Marine Fisheries Service, NOAA. 
4 Wakefield, W. W., II. 2015. Unpubl. data. Newport Facil- 
ity, Northwest Fish. Sci. Cent., Natl. Mar. Fish. Serv., 2032 
SE OSU Dr., Newport, OR 97365-5275. 
well as attributes of school structure indicative of their 
behavioral state. 
Target tracking methods have been used previously 
to quantify changes in fish behavior in several settings. 
For laboratory studies, stereo photography has been 
used, with both target tracking and other analytical 
techniques, to quantify the movement of individual fish 
and the coordinated movement of fish schools (Tien et 
al., 2004; Viscido et al., 2004). For in situ studies, sev- 
eral acoustic methods have been used for target track- 
ing. These methods include the use of buoy-mounted, 
split-beam sonars to track the avoidance of fishing 
vessels or bottom trawls by individual cod (Handegard 
and Tjpstheim, 2005) and the use of DIDSON imaging 
sonars to track the escapement of individual walleye 
pollock ( Gadus chalcogrammus ) from a pelagic trawl 
(Williams et al., 2013). Optical methods are rarely used 
for aquatic in situ experiments because 1) water clar- 
ity typically limits the feasible range that cameras can 
image objects and 2) artificial lighting, which can itself 
alter fish behavior, is needed in other than very shal- 
low water. 
Our initial intention was to use optical target track- 
ing on several species of snappers and groupers that 
are relatively solitary, benthic-oriented, and com- 
monly seen in the study area. However, at one loca- 
tion, a school of vermilion snapper ( Rhomboplites au- 
rorubens ) was viewed repeatedly by the benthic stereo 
cameras, and, during one pass of the towed camera 
vehicle (Lembke et al., 2013), the school was viewed 
during the entire period between the transits of the 
tow vessel and the camera vehicle. Here we focus on 
this single observation of a schooling species rather 
than on our more numerous observations on solitary 
species because our intent is to demonstrate that in 
situ optical target tracking with stereo cameras can be 
used to quantify changes in individual and group be- 
haviors that are often associated with predator avoid- 
ance (Parrish et al., 2002; Viscido et al., 2004) are also 
displayed in response to a camera vehicle. To aid the 
development of stealthier camera vehicles, we consider 
the linkages between these behavioral changes and the 
stimuli produced by the camera vehicle and its tow ves- 
sel that could have triggered the changes. In addition, 
we provide an example of how target tracking can be 
used to estimate sampling efficiency of camera vehicles. 
Materials and methods 
Description of the study 
The experiment was conducted from 17 through 19 July 
2014 on the Florida Middle Grounds, which is a reef 
of calcareous sponges and soft corals situated in the 
northeast Gulf of Mexico (Coleman et al. 1 ; Wakefield 4 ). 
This reef has a relatively flat top, at a depth range of 
23-32 m, and steep slopes. The highest fish abundance 
and diversity concentrated near the junction of the reef 
top and slope. The target species, vermilion snapper, is 
