Stienessen et al.: Comparison of model types for prediction of seafloor trawlability in the Gulf of Alaska 187 
€ 
© 
oO 
- 
x 
= 
io») 
£ 
x= 
—] 
pas 
oO 
Zz 
3.590 3.595 
3.600 3.605 3.610 3.615 
Depth (m) 
3.620 3.625 
Easting (1 x 10° m) 
Figure 1 
An example of bathymetry from a fine-scale multibeam survey conducted with a 
Simrad ME70 echo sounder and examples of the positions of associated camera stations 
deployed in the summers of 2011, 2013, and 2015 in the Gulf of Alaska. Data from mul- 
tibeam surveys and camera deployments were used in models to characterize the extent 
of trawlability of habitat utilized by rockfishes (Sebastes spp.). The double-headed arrow 
indicates the depth-dependent width of the swath of seafloor insonified (1.e., area tar- 
geted) with the Simrad ME70. The black sinuate lines depict the track of the camera as 
it drifted along the seafloor (i.e., camera stations). The blue shading around each of the 
black lines indicates the area within 20 m of a camera station, and the data collected 
within this shaded area with the Simrad ME70 were used to calculate the seafloor char- 
acteristics used in the models. This image shows the north-south transects of 1 fine- 
scale multibeam survey and tracks at 3 camera stations. 
single digital camera included 5 min on the seafloor, and 
each deployment of the SDC system included 15-30 min 
on the seafloor in 2011 or 30 min just above the seafloor in 
2013 and 2015. The SDC system had a real-time video feed 
to the surface and a winch that allowed the camera posi- 
tion to be adjusted vertically in the water column as the 
camera drifted over rough terrain. 
Data extraction and analysis 
Multibeam data Bathymetry and backscatter data were 
extracted from the raw files retrieved from the Simrad 
ME70. More specifically, acoustic power associated with 
each detection of the seafloor was converted to backscat- 
ter following protocols established by Weber et al. (2013) 
in which system gains, calibration offsets, water column 
spherical spreading and absorption, and the area tar- 
geted by the beam are considered. Detections of the bot- 
tom were further used to calculate values for 3 seafloor 
characteristics identified by Pirtle et al. (2015) as among 
the best metrics with the best scale of analysis to predict 
seafloor trawlability in the GOA. The characteristics were 
S;, oblique, VRM, and BPI. The Sj, oblique data were lim- 
ited to the angle-dependent S,, data collected at incidence 
angles between 35° and 50°. 
On the basis of results of Pirtle et al. (2015), VRM was 
calculated with data from a 21-by-21 array of 6-m? grid 
cells by using the bathymetry of each cell and 8 surround- 
ing neighbor cells, and BPI was calculated with data from 
analysis windows, each with a radius of 200 6-m” grid 
cells. In our study, VRM was calculated by using only met- 
ric units (i.e., positions in meters of x and y coordinates in 
the Universal Transverse Mercator system and depth in 
centimeters). In contrast, Pirtle et al. (2015) used 2 differ- 
ent types of units for VRM calculations (1.e., positions in 
degrees of x and y coordinates in a geographic coordinate 
system and depth in meters). 
Only seafloor characteristics calculated for the area 
within 20 m of a given camera station (i.e., within 20 m 
of the path of the camera as it drifted over the seafloor) 
were associated with that particular station. Specifically, 
if the bottom detection was within 20 m of the camera sta- 
tion, VRM and BPI were calculated, even if part or all of 
the array of 6-m? grid cells (VRM) or of the radius of the 
