Zwolinski et a!.: Distributions and abundances of Sardinops sagax and other pelagic fishes in the California Current Ecosystem 
113 
species if catches were less than 75 fish, or, otherwise, 
of a subsample of 50 fish. On four occasions during the 
2010 survey aboard FV Frosti, the floats were removed 
and the net was set on midwater targets. The net was 
directed to the depth of schools with the aid of a Scan- 
mar trawl eye net sounder (Scanmar AS, Asgardstrand, 
Norway). These trawls were largely unsuccessful as fish 
avoided the gear and catch numbers were low to none. 
Data analysis 
Echoes numerically classified as CPS, according to the 
multifrequency algorithm described in Demer et al. (2012), 
were integrated vertically from 3 m below the transducer 
to 70 m depth and were averaged horizontally over 100-m 
intervals along the survey track. The resulting nautical 
area scattering coefficients (s A , m 2 nmi ~ 2 ) at 38 kHz were 
indexed in space and time. Because most pelagic fish 
schools disperse and ascend above the transducer depth 
during night (Cutter Jr. and Demer, 2008), the nighttime 
acoustic samples were considered negatively biased for 
CPS and where not used for abundance estimation. Cells 
sampled during the day, defined here as the time between 
nautical twilights, had their s A apportioned to each 
target species on the basis of the proportion and sizes 
of the species in the nearest trawl (Demer et al., 2012). 
Biomass and numerical densities were obtained from 
the species-apportioned s A , as detailed in Demer et al. 
(2012), by using estimates of average target strength 
(TS; dB re 1 m 2 kg" 1 ; (Barange et al., 1996). Occasion- 
ally, echoes ascribed to CPS were not matched with CPS 
catches. These echoes were often semidemersal, i.e., in 
contact with the seabed, or in conditions unsuitable for 
CPS. Therefore, these echoes were likely from other 
swimbladdered fish species such as hake (also named 
Pacific whiting; Merluccius pi'oduetus ) or rockfishes (Se- 
bastes spp.), which tend to reside deeper than CPS, par- 
ticularly during the day (Dorn et al., 1994; Butler et al., 
2003). These echoes, comprising a small fraction of the 
total s A initially ascribed to CPS, were excluded from 
further analysis. 
The spatial match between the acoustically detected 
CPS and the trawl catches was tested by resampling. 
First, the s A attributed to CPS during daytime was av- 
eraged for spatial bins with various sizes, each centered 
on the locations of catches with CPS. These values 
represent the average s A associated with CPS catches. 
Then, 1000 sets of points with equal number to those 
of the CPS catches were drawn randomly within the 
daytime transect track. The mean acoustic backscatter 
in the vicinity of those points represents the average s A 
ascribed to CPS in the total survey field. The 95% con- 
fidence intervals (CI 95 ) for the average s A of CPS were 
chosen from the resampled distribution according to the 
percentile method (Efron, 1981). Association between 
CPS catches and acoustically detected CPS is strongly 
supported when the s A ascribed to CPS in the vicin- 
ity of CPS catches is above the confidence intervals. 
The potential habitat of sardine in the CCE, defined 
here as the region expected to contain an average of 
90% of all adult sardine (Zwolinski et al., 2011), was 
predicted for each survey period with a generalized 
additive model (GAM). The model is based on a 12-year 
time series of pump-sampled sardine-egg presence and 
satellite-sensed measurements of sea-surface tempera- 
ture, chlorophyll-a concentration, and the gradient of 
sea-surface height. The distributions of sardine esti- 
mated from the acoustic-trawl surveys were visually 
compared to those of their potential habitat. 
For each target species, the survey area was then 
stratified into one or two strata with comparable bio- 
mass densities and approximately equal transect spac- 
ing. To the north and south, the strata extended beyond 
the exterior transects by half of the intertransect dis- 
tance. To the east and west, the strata were bounded by 
the coastline and by the offshore limits of the transects. 
Occasionally, the inshore and offshore limits were de- 
fined by lines parallel to the coast, by excluding large 
areas of zero densities and ensuring uniformity in the 
length of the transects within the strata. Mean biomass 
densities for each strata were obtained by a transect- 
length-weighted average of the mean transect densi- 
ties, which is equivalent to the arithmetic mean of all 
integration cells in each strata. Total biomasses were 
estimated by multiplying the mean biomass densities 
by the areas of the respective strata and by summing 
across strata. Confidence interval (CD and coefficient 
of variation (CV) values for total abundance were es- 
timated from bootstrap resampling (Efron, 1981) of 
the mean biomass densities of the transect means, as 
described in Demer et al. (2012). Statistical indepen- 
dence between the transect means, required to provide 
unbiased estimates of variance, was verified for every 
species and strata through an autocorrelation analysis. 
Further details of the data processing are provided in 
Demer et al. (2012). 
Sardine-length distributions were estimated by a 
weighted average of the length distributions from the 
individual trawls, by using as weights the sardine den- 
sities estimated with the nearest acoustic samples. The 
latter were obtained by converting s A values into nu- 
merical densities with individual TS-to-length equations 
(Barange et al., 1996). Total sardine numbers were ob- 
tained by summing abundances across lengths. Because 
no new recruits were visible in the time series, the net 
instantaneous mortality rate of the stock was estimated 
by fitting an exponential-decay function to the measures 
of total sardine abundance versus time. For sardine, the 
only actively managed species distributed throughout 
most of the survey area, the survey estimates of abun- 
dance, demographics, and mortality were compared with 
those from the independent assessment. 
Results 
CPS distributions 
The results from the multifrequency algorithm provided 
evidence that CPS were abundant and broadly distrib- 
