222 
Fishery Bulletin 119(4) 
SS 
Results 
General observations 
The array of 3 MOUSS platforms was deployed during 
daylight hours at 13 distinct areas over 9 d in August 
2014 and at 7 distinct areas over 5 d in July and August 
2015. Deployments of MOUSS platforms varied in time 
from 7 to 10 h. Over the 9 d of dedicated sampling in 
2014, 73 transects were surveyed with vehicles, resulting 
in 218 individual passes or observations in front of the 
MOUSS platforms. Over the 7 d of dedicated sampling in 
2015, 57 transects were surveyed with vehicles, resulting 
in 171 individual passes or observations in front of the 
MOUSS platforms. 
The target speed of the AUV was 1.0 kt (0.51 m/s), and 
target altitude was either 2 or 4 m. In 2014, to improve 
the likelihood of observing the AUV in images from the 
stereo camera, the transect altitude was lowered from 
4 to 2 m. The AUV transited in straight lines but was 
vulnerable to heavy currents, which caused the vehicle to 
“crab,” make frequent course corrections, and to transit 
at a more variable rate of speed. The target speed of the 
TV was approximately 3.5 kt at an altitude of 2 m, but 
sometimes the speed was as high as 5 kt when travelling 
down current and the altitude was as high as ~5 m in high- 
relief areas where the potential of hitting the reef was a 
concern. Because the TV is towed and control of the vessel 
is fairly precise at the surface (e.g., GPS navigation), the 
TV transited the straightest transects of all the vehicles. 
However, transect deployments had to be well planned and 
executed in order to avoid hitting obstacles (e.g., tending 
buoys or reef). The target speed and altitude of the ROV 
was 0.5—1.0 kt and 0.5—-1.0 m above the seafloor, but some- 
times speed was as high as 2 kt and altitude was as high 
as 2.5 m because of currents, tether orientation, and oper- 
ator error. Currents, tether orientation, and pilot error 
sometimes caused the transect path to meander from the 
intended straight line. In general, the ROV was the vehicle 
most consistently observed on MOUSS cameras, and tran- 
sects improved over time likely as a result of a combination 
of slow speeds and rapid feedback to the pilot. 
Generalized additive models 
Results from the fish acclimation GAMs indicate signifi- 
cant effects for both time and habitat complexity. The 
model explained 17.8% of deviance with a coefficient of 
determination (r”) of 0.11. Although the information on 
model fit indicates that there is considerable unexplained 
variability, the trend indicates that fish counts declined 
during the acclimation period and reached an asymptote 
at around 45 min (Fig. 4). General observations indicate 
that jacks (Seriola spp.) and great barracuda (Sphyraena 
barracuda), when present, closely examined the station- 
ary camera but lost interest quickly. In addition, the reef 
fish community observed immediately around the stereo 
cameras became more active at initial deployment but 
appeared to slowly become disinterested and begin to 
° 
to 
ou 
Standardized count 
[o) 
[@) 
[@) 
Minutes 
Figure 4 
Changes in fish abundance during the acclimation period 
(first 60 min) following deployment of Modular Optical 
Underwater Sampling System platforms at selected reef 
sites in the Florida Middle Grounds. The gray shaded area 
represents the 95% confidence interval. Trends indicate a 
decrease in abundance with an asymptotic response occur- 
ring at around 45 min. The experiment was conducted 
during August 2014 and July and August 2015. 
forage and interact with the habitat and each other at 
around 30 min after deployment. 
Results from vehicle interaction GAMs indicate that 
significant terms and overall fit of the models differed 
between behavioral guilds and between vehicles within 
each guild (Table 2). From a qualitative perspective, the 
pelagic pursuers group had strong attraction to all of the 
vehicles tested and was the most obvious group to charac- 
terize (Fig. 5). The model for the AUV had significant effects 
for RVA, vehicle range, and transect number, but habitat 
complexity was not significant. The AUV model explained 
48.9% of the deviance with an r” of 0.43. The TV model 
had significant effects for vehicle range, transect number, 
and habitat complexity, but RVA was not significant. The 
TV model explained 59.3% of the deviance with an r” of 
0.29. The ROV model showed significant effects for vehicle 
range, but vehicle altitude, transect number, and habitat 
complexity were not significant. The ROV model explained 
81% of the deviance with an r” of 0.46. Pelagic pursuers 
had spikes in abundance when vehicles were close in prox- 
imity to the MOUSS platforms (~50 m) (Fig. 5). Passage 
of the slower-moving AUV and ROV resulted in a spike in 
relative fish abundance at closer ranges (~10 m) than pas- 
sage of the fast-moving TV (~50 m). Relative abundance of 
pelagic pursuers increased with increasing AUV altitude 
but had no relationship to TV altitude. The effect of rela- 
tive altitude could not be tested on the ROV because all 
passes were within the low category. Increasing the num- 
ber of transects generally resulted in peak increases in 
relative abundance between the third and fourth transect 
for the AUV and TV, but this pattern was not evident for 
the ROV. 
