Bacheler and Shertzer: Estimating relative abundance and species richness from video surveys of reef fishes 
21 
~ 0.04 
3 
o 0 02- 
c 
03 
^ o.oo- 
LU 
cc 
- 0 . 02 - 
-0.06- 
A 
0.06- 
B 
0.06- 
T 
0.04- 
-r- 
0.04- 
| — r— _ 0.02- 
T -r- 0.02- 
1 1 1 ' 1 1 ' 1 1 ' | n no 
□ □r 0 00- 
r~l l l “ uuu 
_i_ - 1 - ^ -0.02- 
-0.02- 
4- 
-0.04- 
-0.04- 
-0.06- 
-0.06 - 
“I T 
25 50 100 
200 
"T T 
25 50 100 
200 
I 1 1 I I 
25 50 100 200 
n 
a; 
ra 
o 
v) 
> 
O 
6 - 
4- 
D 8- 
_ M 
00 
L 
I 6- 
4 - 
| 
1 4- 
V 
2 - 
2 - 
- — 
r--.- 
0- 
0- 
50 100 150 200 
Number of frames 
50 100 150 200 
Number of frames 
I 
50 
T 
0 50 100 150 200 
Number of frames 
Figure 3 
Top row: mean relative error (MRE) of MeanCount, the mean number of individual fish observed in a series 
of frames over a viewing interval, across all videos analyzed in this study from the northern Gulf of Mexico 
in 2001-2002 and 2004-2007, as a function of the number of video frames read for (A) red snapper (Lutja- 
nus campechanus), (B) vermilion snapper ( Rhomboplites aurorubens), and (C) scamp ( Mycteroperca phenax). 
Boxes represent the interquartile range, thick solid lines represent medians, and whiskers extend to the 
most extreme data point within 1.5 times the interquartile range from the box. Bottom row: coefficient of 
variation (CV) of MeanCount as a function of the number of video frames read for ( D ,) red snapper (Lutjanus 
campechanus ), (E) vermilion snapper (Rhomboplites aurorubens), and (F) scamp (Mycteroperca phenax ). In 
each panel, curves represent CVs from each sampling event (i.e. , each 20-min video collection), computed from 
1000 bootstrap replicates. Each CV curve is scaled to its minimum. 
in 25 frames on the basis of the mean duration of 
each species in a video (estimated degrees of freedom 
[edf] = 1.5; F= 36.6; PcO.0001), mean number of indi- 
viduals in a video (edf=1.0; F= 43.9; P<0.0001), or their 
interaction (edf=7.8; F= 0.7; P=0.004). Species were ob- 
served with higher probability as their mean time in 
the videos increased; however, this probability saturat- 
ed near 1.0 for mean times of 100 s or more (Fig. 6A). 
Similarly, the probability of being observed increased 
as the mean number of individuals increased, but, the 
trend was nearly linear over the range of the predictor 
(Fig. 6B). The families of fishes that were most likely 
to be observed in 25 frames of video were the generally 
sedentary groups like jawfishes, bigeyes, squirrelfishes, 
angelfishes, and triggerfishes, and those families most 
likely to be missed were fast-moving groups like tunas 
and mackerels, barracudas, and jacks (Table 2). 
Discussion 
In many places around the world, underwater video 
has become a common approach to monitor the abun- 
dance and distribution of marine fish and invertebrate 
species and to quantify marine biodiversity (e.g., Heag- 
ney et al., 2007; Stobart et al., 2007; Brooks et al., 
2011; Merritt et al., 2011; Gladstone et al., 2012). For 
many such studies, BRUVS have been used and have 
provided an index of the abundance of various species 
through the use of a stationary point-count with the 
MinCount method (Ellis and DeMartini, 1995; Willis 
et al., 2000; Murphy and Jenkins, 2010). Recent re- 
search has indicated that MeanCount is more linearly 
related to true abundance than is MinCount (Conn, 
2011; Schobernd et al., 2014). To provide the next logi- 
cal step in the evaluation of the MeanCount approach, 
