von Szalay and Somerton: A method for predicting trawlability in the Gulf of Alaska 
499 
Table 1 
Settings of the bottom-line-picking algorithm in Echo- 
view software used before classification of seabed habi¬ 
tats. These values “told” the software how to detect 
the bottom (pick the bottom line) 
in the acoustic data 
collected during the National Marine Fisheries Service 
biennial bottom trawl survey of the Gulf of Alaska. 
Setting 
Value 
Start depth (m) 
10 
Stop depth (m) 
500 
Min S v for good pick (dB) 
-80.00 
Discrimination level 
-40.00 
Back step range 
0.00 
Peak threshold (dB) 
-40.00 
Minimum threshold (dB) 
-70.00 
tative. Next, the Echoview algorithm for background 
noise removal was applied to the data by using a sig- 
nal-to-noise ratio setting of 10 and a maximum noise 
level subject to a removal setting of 0 dB. Echoview 
algorithms for bottom classification were then used to 
derive the 9 standard feature parameters from the first 
and second echo returns of the signal from the bottom. 
Echoview Software has determined that these feature 
parameters distinguish general seabed features. No 
attempt was made to narrow the 9 features down to 
those believed to be most relevant in distinguishing 
trawlability. The bottom echo threshold at 1 m, which 
is used to determine the end of the first and second 
echoes, was set to -125 dB and the reference normal¬ 
ization depth was set to 300 m. 
The names of the feature parameters derived from 
the first echo are: roughness, first bottom length, bot¬ 
tom rise time, depth, maximum S v (max S v ), kurtosis, 
and skewness; the names of the parameters derived 
from the second echo are: hardness and second bottom 
length (Table 2). Information on how Echoview process¬ 
es backscatter data, including definitions of the terms 
used to derive the 9 feature parameters, and equations 
defining the parameters can be found in the help file 
for the software (available from website). Each of the 
238 acoustic data segments consisted of ~50 records, 
where a record is made up of the 9 acoustic feature pa¬ 
rameter values derived from groups of 10 consecutive 
pings. In this study, a record is the basic classification 
unit of an acoustic segment. The parameter values of 
the individual records were modeled directly without 
using the classification feature in the Echoview bottom 
classification module, which uses principal component 
and cluster analyses to categorize individual records 
into a user-specified number of bottom types. 
Trawlability was modeled as a function of the 9 acous¬ 
tic feature parameters by using both generalized linear 
modeling and generalized additive modeling functions 
in R, vers. 3.2.0 (R Core Team, 2015). A binomial error 
distribution was assumed for both types of models. 
Three different classes of generalized linear models 
(GLMs) were evaluated by using the minimum value of 
the Akaike’s information criterion (AIC) to choose the 
best-fitting model within each class (Table 3). The 3 
classes consisted of models with all linear terms, mix¬ 
tures of linear and polynomial terms, and mixtures of 
Table 2 
Definitions of the 9 standard parameters used in the Echoview algorithms for classification of seabed habitats. 
These feature parameters were covariates in all models used in this study to predict whether sampling areas 
in the Gulf of Alaska were 
trawlable. X=affirmative. 
Depth 
Parameter 
Definition 
normalized 
First echo 
Roughness 
Tail energy. Integration of the tail of the first echo. Exclusively due to 
incoherent backscattering from facets inclined towards the transducer 
X 
First bottom length 
Total duration of the first bottom echo (bottom line depth to bottom 
echo threshold at 1 m). 
X 
Bottom rise time 
Depth 
Max Sv 
Kurtosis 
Skewness 
Attack duration (bottom line sample to the peak sample of the first echo). 
Water depth 
Maximum energy. Maximum volume backscatter strength 
Tailedness. Sharpness of the first echo peak 
Asymmetry around the first echo peak 
X 
Second echo 
Hardness 
Total energy of the second echo (integration of the complete second 
acoustic bottom return). 
X 
Second bottom length 
Total duration of the second echo (bottom line depth to bottom echo 
threshold at 1 m). 
X 
