498 
Fishery Bulletin 11 5(4) 
the rate and precision at which the bottom-trawl sur¬ 
vey sampling cells are classified to improve our ability 
to efficiently sample areas that are trawlable and avoid 
untrawlable ground. One approach described by Weber 
et al. (2013) and Pirtle et al. (2015), involves modeling 
trawlability as a function of seafloor metrics derived 
from bathymetry and backscatter data collected from 
multibeam acoustic surveys conducted by the National 
Marine Fisheries Service. Because no multibeam acous¬ 
tic data are collected during the GOA trawl survey, we 
considered another method for this study, one that is 
based on the analysis of calibrated, single-beam acous¬ 
tic backscatter data. The Alaska Fisheries Science Cen¬ 
ter has routinely collected these data since 2005 on all 
chartered fishing vessels used to conduct the bottom 
trawl survey. Although single-beam acoustic backscat¬ 
ter from the seabed has been analyzed in a number of 
studies to distinguish a variety of habitat types (Kloser 
et al., 2001; Anderson et al., 2002; Freitas et al., 2003; 
Riegl et al., 2005; Bartholoma, 2006), we consider the 
response variable as binary, as simply distinguishing 
between trawlable and untrawlable bottoms. The ob¬ 
jectives for this study were 1) to examine the feasibil¬ 
ity of developing a trawlability prediction model based 
on backscatter data from areas of known trawlability 
and 2) to evaluate the use of applying the model to 
predict trawlability in unknown areas on the basis of 
measured backscatter properties of acoustic data col¬ 
lected along track lines of future surveys conducted by 
the same vessel and with the same echosounder that 
was used to collect data for this study. 
Materials and methods 
The acoustic data were collected aboard the 38.4-m 
stern trawler FV Sea Storm during the 2013 GOA bot¬ 
tom trawl survey by using a Simrad 1 ES60 echosound¬ 
er (Kongsberg Maritime AS, Horten, Norway) equipped 
with a 7.1° beam width, 38-kHz, split-beam transducer, 
which operated at a ping rate of 1 Hz and pulse du¬ 
ration of 1.024 ms. The echosounder was calibrated 
on-axis with a copper sphere according to standard 
procedures described by Foote et al. (1983). A total of 
238 individual acoustic data segments were used in the 
analysis, half from trawlable and half from untraw¬ 
lable areas. The trawlable and untrawlable segments 
were randomly selected from a pool of segments span¬ 
ning the entire range of the survey area that satisfied 
the basic criteria identified in the next paragraph for 
the sampling cells containing the segments (Fig. 1). 
An acoustic data segment is an echosounder-insonified 
section of a vessel track line, and all segments from 
both trawlable and untrawlable areas consisted of data 
collected over 15-min time intervals, corresponding to 
the duration of a standard trawl haul. Although the 
1 Mention of trade names or commercial companies is for iden¬ 
tification purposes only and does not imply endorsement by 
the National Marine Fisheries Service, NOAA. 
depth range of the biennial bottom trawl groundfish 
survey extends to water depths as deep as 1000 m, the 
depths associated with the segments of this specific 
study were all less than 300 m because unacceptably 
slow ping rates (producing poor echogram resolution of 
the seabed) are required for deeper depths. Depths less 
than 300 m comprised 90% of the survey area (Fig. 1). 
As with the selection process for trawlable and un¬ 
trawlable segments, the sampling cells containing the 
selected segments were chosen randomly from a pool 
of cells satisfying certain criteria. Among the trawlable 
cells, only sampling cells that represented areas that 
had been successfully towed without any documented 
incidents such as tears in nets or trawl door entangle¬ 
ments with the bottom on at least 2 separate surveys 
were included in the analysis. Among the untrawlable 
cells, only sampling cells classified as untrawlable owing 
to 1 of the 5 hard or rough categories, or combination 
categories (i.e., hard+rocky, rolling seabed, pinnacles, 
snags, ledges) were used in the analysis. Cells classi¬ 
fied as unnavigable were not used because acoustic data 
cannot be collected from areas that the survey vessel 
cannot navigate. Furthermore, the fixed fishing gear 
and underwater cable categories are for cells with man¬ 
made obstructions; these cells do not necessarily have 
acoustic signatures that identify them as untrawlable, 
yet it would be ill advised to trawl in these areas. An¬ 
other major reason for considering an area untrawlable 
is that it is considered too steep. However, for the pur¬ 
poses of this analysis, such cells were not considered 
because acoustic features associated with steep slopes 
have been shown to be distinct from those of more level 
areas, regardless of substrate type (von Szalay and Mc- 
Connaughey, 2002). Furthermore, steep slope areas are 
primarily confined to relatively deep waters (>200), and 
the models developed in this study are intended only 
for use in the continental shelf portion of a survey area. 
The raw acoustic data files were processed before 
analysis to remove noise in the form of triangle wave 
dither that degrades the ES60-generated raw files, by 
using the known period and amplitude of the dither 
(Ryan and Kloser 2 ). The triangle wave-corrected raw 
files were subsequently analyzed by using the seabed 
classification module in Echoview, vers. 6.1.72 (Echo- 
view Software, Pty. Ltd., Hobart, Australia). Seven set¬ 
tings, which are used by an algorithm within the soft¬ 
ware to detect the bottom by using the data collected 
(bottom line pick), were specified before we derived 
classification data. The values of the settings used in 
this study (Table 1) were the defaults recommended by 
Echoview under most circumstances, except for the val¬ 
ue for the minimum volume backscatter strength (min 
S v for good pick), which was modified for this study 
after consulting with an Echoview Software represen- 
2 Ryan, T., and R. Kloser. 2004. Quantification and correction 
of a systematic error in Simrad ES60 echosounders, 9 p. 
ICES FAST, Gdansk. [Available from Marine and Atmospher¬ 
ic Research, Commonwealth Scientific Industrial Research 
Organisation, GPO Box 1538, Hobart, TAS 7001, Australia.] 
