90 
Fishery Bulletin 11 7(1-2) 
see the trap. Staying at least 15 m from the trap, div¬ 
ers counted fish and measured habitat features along 
3 transects that were each 30 m long and 10 m wide 
(i.e., they surveyed 300 m 2 for each transect and 900 
m 2 at each site). The chosen direction of the first tran¬ 
sect was typically toward the part of a site that had 
the most hard-bottom substrate, beginning ~15 m from 
the trap, so that the trap and video sampling were dis¬ 
turbed as little as possible. The second transect was 
surveyed in the direction opposite of the first transect, 
and the third transect was perpendicular to the first 
2 transects and often contained the least amount of 
hard-bottom habitat. The depths of habitats sampled 
by UVC were similar (with differences <2 m) in all 
cases to the depths at which traps were deployed and, 
therefore, the depths at which video recordings were 
made. Further details about the method and full re¬ 
sults of UVC can be found in Bacheler et al. (2017). 
For the work described in this manuscript, the pri¬ 
mary interest in UVC data was whether divers encoun¬ 
tered a shark at a particular site and, if they did, the 
recorded species identification for the sharks observed. 
We compared the identification of sharks by divers with 
that from video analysis at those sites where observa¬ 
tions were positive for both survey methods. We used 
a chi-square test to compare the proportion of sites at 
which sharks were seen by divers with the proportion 
of sites for which sharks were seen in video analysis. 
All statistical analyses were conducted with SigmaPlot, 
vers. 11.0 (Systat Software, Inc., San Jose, CA). 
Results 
Species of sharks observed in videos 
We reviewed 100.5 h of video and observed sharks in 
videos from 30 of 77 sites. In 5 cases, pairs of sites 
were located <500 m apart, and review of videos from 
these sites revealed that the same species of sharks 
were observed at both sites in each pair. Therefore, 
video recordings from one of the pair of sites (a total 
of 5 videos) were dropped from further analyses. As a 
result, 25 of 72 sites with shark observations, or 35% 
of videos, were used for analyses. The average length 
of video recordings was 78.5 min (SD 6.00). For most 
sites, a single species of shark was observed, as a soli¬ 
tary individual in video frames. However, for 2 sites, 
2 nurse sharks (Ginglymosto?na cirratum) were seen 
at the same time in a single video frame, and 2 At¬ 
lantic sharpnose sharks (Rhizoprionodon terraenovae ) 
were seen in a single video frame from another site. 
In each of 18 videos, 1 species of shark was observed; 
6 videos contained 2 species, and 1 video contained 4 
species. In all of these analyzed videos, a minimum of 
36 individual sharks were observed. At the 25 sites 
for which sharks were observed in videos, their as¬ 
sociated chevron traps contained significantly more 
fish, with a mean of 71.9 individuals (SD 56.5), than 
those traps associated with the videos in which sharks 
were absent, with a mean of 48.4 individuals (SD 49.9) 
(72=47 sites; Mann-Whitney U test: U- 397, P=0.03). No 
sharks were captured in traps, and the full results for 
trap catches can be found in Bacheler et al. (2017). 
Seven species of sharks, including the nurse, tiger 
(Galeocerdo cuvier), spinner (Carcharhinus brevipin- 
na), sandbar (C. plumbeus ), Atlantic sharpnose, bull 
(C. leucas ), and lemon (Negaprion brevirostris) sharks, 
were observed in videos (Fig. 2). Because of water tur¬ 
bidity and proximity of sharks to the camera, we were 
unable to identify to species carcharhinid sharks ob¬ 
served in videos from 2 sites. In most cases, the sharks 
that we observed were readily identifiable to species. 
However, one limitation of any method of analyzing 
video to sample fish is that taxa are not retained for 
positive identification. 
Frequency of occurrence (Fig. 3) was greatest for 
the nurse shark, which was observed in videos from 
-14% of sites (10 of 72 sites), followed by the tiger (8%, 
6 sites), spinner (7%, 5 sites), sandbar (5%, 4 sites), 
and Atlantic sharpnose (4%, 3 sites) sharks. The bull 
shark, lemon shark, and sharks identified only to the 
family Carcharhinidae were observed in videos from 
3% of sites (2 of 72 sites). 
Comparison of methods 
When the entire video was analyzed, sharks were 
observed at 4 times more sites than when only the 
SERFS 20-min segment was analyzed (28% versus 7% 
of the total number of sites, respectively; chi-square 
test: x 2 =9.5, df=l, P=0.002). Sharks frequently were 
observed in videos classified as outside of the 20-min 
segment. Of the 25 videos with sharks, 20 videos con¬ 
tained sharks exclusively outside of the 20-min seg¬ 
ment, and 5 videos had sharks within the 20-min seg¬ 
ment. Of these 5 videos, 3 videos included sharks that 
also were observed later in the video outside the 20- 
min segment. Of the 20 videos with sharks observed 
solely outside the 20-min segment, 2 videos had sharks 
before the 20-min segment (10%), and 18 videos con¬ 
tained sharks after the segment (90%). 
Results from analyses of entire videos for sharks 
were superior to those based on UVC of transects by 
divers. A greater diversity of sharks were observed and 
sharks were recorded at a larger number of sites when 
the alternate video-analysis method was used than 
when divers used UVC. At 5 sites, sharks were de¬ 
tected both by using the SERFS 20-min video segment 
and by using UVC, although the individual sites where 
sharks were detected were mostly (except for 1 site) 
distinct between methods. By using UVC, 3 species of 
sharks were recorded at 5 sites (7% of sites, Table 1), 
and the use of video analysis resulted in observations 
of 7 species in videos from 25 sites (35% of sites; chi- 
square test: x 2 =15.2, df=l, P<0.001). Although a greater 
diversity of sharks overall was observed by using video 
analysis than by using UVC, site-specific comparisons 
of UVC and video data revealed differences in the pres¬ 
ence and absence of sharks, as well as in the species of 
