56 
Fishery Bulletin 110(1) 
potential habitat, to the north in the spring and to the 
south in the summer. This coverage served to assure 
that most of the sardine were indeed located within the 
predicted potential habitat. 
Within their habitat, the dominant CPS in the CCE 
may be broadly and patchily distributed (Table 1), but 
are mostly aggregated in clusters of dense schools (Cut- 
ter Jr. and Demer, 2008; McClatchie, 2009). Because 
sampling of such skewed distributions is often the larg- 
est source of variance in acoustic-trawl surveys (Pen- 
nington, 1983; Demer, 2004), echosounder sampling was 
conducted continuously, and trawl sampling was con- 
ducted periodically, along parallel-line transects (Fig. 
2). A large intertransect distance allowed independence 
of the mean acoustic backscatter between transects, 
permitting statistically unbiased estimations of mean 
biomass densities and sampling variances for target 
species (Jolly and Hampton, 1990). Transect indepen- 
dence was tested by applying the auto-correlation func- 
tion to values of integrated echo energy for each species 
and stratum for all intertransect lags (distances). 
Trawl sampling 
CPS generally aggregate into schools during day and 
disperse, mix, and rise to the sea-surface at night (Hol- 
liday and Larsen, 1979; Cutter Jr. and Demer, 2008). 
Therefore, trawl sampling for identifying species and 
their sizes was performed at night, either at uniformly 
or randomly distributed, pre-assigned, or ad hoc stations 
along the transects. The trawl used was a Nordic 264 
rope trawl with an opening of 600 m 2 (NET Systems, 
Bainbridge Island, WA). To minimize fish-size selectiv- 
ity, the codend was fitted with an 8-mm-square mesh 
liner. The headrope was rigged with floats for towing at 
the surface at a speed of nominally 3.5 kn for 30 min. Up 
to four trawls were performed each night, beginning 30 
to 60 min after sunset. The catch was sorted by species 
and weighed. From the catches with CPS, up to 75 fish 
from each target species were randomly selected. Those 
were weighed (g), and measured (mm), either in stan- 
dard length (SL) for sardine, anchovy, and herring, or 
fork length (FL) for jack mackerel ( Trachurus symmetri- 
cus) and Pacific mackerel. The length distributions of the 
sampled populations were estimated by using weighted 
averages of the length distributions from the trawls. The 
length data were first combined by transect, weighted 
by the acoustically estimated mean densities closest to 
each trawl. Then, the transect-weighted lengths were 
combined, weighted by the acoustically estimated mean 
densities for each transect. 
Echosounder sampling 
Echosounder sampling was conducted by using multi- 
frequency (18, 38, 70, 120, and 200 kHz) transceivers 
(Simrad EK60; Kongsberg, Norway) configured with 
split-beam transducers (typically Simrad ES18-11, 
ES38B, ES70-7C, ES120-7C, and ES200-7C, respec- 
tively). The echosounder systems were calibrated before 
each survey by using the standard sphere technique 
(Foote et al., 1987) and a 38.1-mm diameter sphere made 
from tungsten carbide with 6% cobalt binder material. 
Throughout the surveys, conducted at a nominal ship 
speed of 10 kn, the echosounders synchronously trans- 
mitted 1024-ps pulses every 0.5 s with powers equal to 
2000, 2000, 1000, 500, and 100 W at 18, 38, 70, 120, 
and 200 kHz, respectively. Following each transmis- 
sion, received-echo power ( p r ; W) data, indexed by time 
and geographic position, were recorded for periods cor- 
responding to an observational depth of 250 m. The 
survey-depth range accommodated the maximum depth 
(70 m) of the expected sardine distribution, and that 
of other CPS (Table 1). With postprocessing software 
(Myriax Echoview; Hobart, Tasmania), the p r values 
were converted to estimates of volume backscattering 
coefficient (s v ; nr -1 ), and volume backscattering strength 
(S^^IO log (s v ); dB re 1 in -1 ). 
Data analysis 
Target identification Echoes may originate from sar- 
dine or other CPS such as jack mackerel, Pacific mack- 
erel, northern anchovy, Pacific herring, and Pacific saury 
( Cololabis saira)\ semidemersal fish such as Pacific hake 
(also called Pacific whiting [Merluccius productus]) and 
rockfishes {Sebastes spp.); and krill (principally Euphau- 
sia pacifica and Thysanoessa spinifera). When analyz- 
ing the echosounder data, it was therefore necessary 
to objectively filter “acoustic bycatch,” i.e., backscatter 
not from the target species. Table 1 summarizes some 
relevant features of bycatch candidates. More detail 
regarding the principal target, sardine, is provided in 
the Appendix. 
Identification of echoes from CPS, i.e., epipelagic 
fishes with swimbladders, was performed with a semi- 
automated data processing algorithm. First, background 
noise was estimated for each echosounder frequency and 
incoherently subtracted from the respective echograms 
of S u . Portions of the “noise-reduced” echograms were 
designated “bad data” if the associated vessel speed was 
below a 5-kn threshold, indicating it was “on station,” 
or otherwise “off effort.” 
Next, the S v values in these “speed-filtered” echo- 
grams were preliminarily identified as echoes from 
fish with swim bladders if their sample-wise variance- 
to-mean ratio (VMR; Demer et al., 2009a) were within 
the -60 dB to -16 dB range. The S v values outside this 
VMR range were set to -999 dB (practically zero). The 
“VMR -filtered” echograms were gridded into ten-sam- 
ple-deep by three-transmission-long bins. The analysis 
bins were smaller than those used in studies of deeper 
dwelling fishes to accommodate the typical dimensions 
and shallower depths of CPS schools. The S v values 
within each depth-distance window were replaced by 
the median value of the S v ensemble. This procedure re- 
duced the variance of the data and allowed comparisons 
of the median S v values with predictions of backscat- 
tering spectra, backscatter versus frequency, for CPS. 
The echograms were ultimately apportioned to CPS, 
