Bacheler and Shertzer: Estimating relative abundance and species richness from video surveys of reef fishes 
17 
Figure 1 
Sampling locations (o) where video was collected during the National Marine Fisheries Service’s reef 
fish video survey in the Gulf of Mexico in 2001-2002 and 2004-2007 as part of its Southeast Area 
Monitoring and Assessment Program. The light gray bathymetric contour lines indicate depths of 50 
and 100 m, respectively. Note that symbols overlap in many cases. 
ing the interval from t=0 to t= 20 min, resulting in anal- 
ysis of 1201 frames. Species were recorded if they were 
listed in the fishery management plans of the Gulf of 
Mexico or South Atlantic Fishery Management Council 
or if commercial or recreational landings were known 
to exist. The resulting list of observed taxa included 
210 species (a group, hereafter, called priority species). 
The time each individual fish swam into and out of 
view (i.e., time in-time out) was recorded for each pri- 
ority species. Our analysis included only those video 
samples in which at least one priority species was seen 
at some point during the 20 min. On the basis of these 
criteria, 1543 videos were included in our analyses. 
MeanCount bias and precision 
For a single video v, the MeanCount of a species across 
video frames was defined with the following equation: 
(EL%) 
MeanCount vF — -, 1 1 ) 
F 
where n f = the number of individuals observed in frame 
f; and 
F = the total number of frames read. 
To examine how MeanCount relates to the number of 
frames read, we chose as case studies 3 ecologically and 
commercially important focal species: red snapper ( Lut - 
janus campechanus ), vermilion snapper ( Rhomboplites 
aurorubens), and scamp (Mycteroperca phenax). These 
species were chosen because they vary substantially in 
terms of body size and schooling behavior. Scamp are 
generally solitary, red snapper often form small groups, 
and vermilion snapper often form large groups. Vermil- 
ion snapper are also much smaller than red snapper 
or scamp. Misidentification of these species is very un- 
likely because of their distinct body shapes and swim- 
ming behaviors. Of the 1543 videos examined in our 
study, red snapper were observed in 375 videos, ver- 
milion snapper in 217, and scamp in 466. MeanCount 
was computed only from those videos in which the focal 
species was observed. 
For each species, the true MeanCount for each video 
(MeanCounty true ) was computed from the full sampling 
universe of 1201 frames. That true value was then esti- 
mated with a subset of frames with a possible sample 
size (F) from the interval [1, 200]. The sampling was 
conducted as follows. First, a list of frames («=200) was 
drawn at random and without replacement from the 
full set of 1201 frames. Then, the first frame of the 
list (F= 1) was used to compute MeanCountyi. Next, the 
second frame (F= 2) was included along with the first 
to compute MeanCount v2 , and so forth until all 200 
frames (F=200) were used to compute MeanCounty^oo- 
We quantified bias and precision in estimates of 
MeanCount with a bootstrap procedure. In the boot- 
strap, the previously described sampling approach 
was repeated 1000 times. That is, for each bootstrap 
iteration b, a new set of 200 frames was drawn to com- 
