Demer et al Seasonal migration of Sardinops sagax in the California Current Ecosystem 
65 
estimates of sardine biomass are not significantly dif- 
ferent from each other or from the assessment-model 
estimates. Also, the spring and summer estimates of 
sardine distribution agree with the model predictions 
of potential sardine habitat (Fig. 3). 
To further minimize the random sampling error in 
these surveys, sampling effort could be increased, or 
modeled predictions of potential sardine habitat could 
be used to optimally plan both DEPM and acoustic- 
trawl surveys, and thus save valuable time to increase 
sampling effort in areas of expected sardine presence. 
These strategies could potentially lower the variance 
and increase the accuracy of the estimates, or allow 
better sampling of the distribution of the less abundant 
species, such as anchovy and Pacific mackerel. Fur- 
thermore, as indicated by the CV values (Table 3), the 
sampling variance could be improved with more trawl 
sampling in the areas with higher acoustically observed 
fish densities (Petitgas et ah, 2003). 
Zwolinski et al. (2011) include a quantitative analy- 
sis of the potential benefits (i.e., reduced effort or de- 
creased estimation variance) of using model predictions 
of potential sardine habitat to optimize sampling. In 
this study, the model from Zwolinski et al. (2011) indi- 
cated that the potential sardine habitat during spring 
2008 was in a region offshore of southern California; 
and during summer of the same year, it was confined 
close to the coast and extended north to Canada. Dur- 
ing both seasons, virtually all of the sardine biomass 
estimated by the acoustic-trawl method was inside 
the “optimal” and “good” habitat. Furthermore, the 
tiny fractions that were mapped outside of the poten- 
tial sardine habitat, to the north in spring and to the 
south in summer, may not have been sardine because 
the closest trawl catches with sardine were at least one 
hundred miles away. 
The results of the acoustic-trawl surveys conducted 
during these two seasons clearly show that the sardine 
were in these predicted habitat regions and therefore 
migrated north between spring and summer. Therefore, 
the benefits described in Zwolinski et al. (2011) should 
be attainable for sardine sampling, irrespective of the 
survey method. If this had been a single-species survey, 
about half of the transects (i.e., north of 38°N) could 
have been reallocated in the spring survey. In summer, 
about one-third of the offshore sampling could have 
been reallocated. Without making many assumptions 
(i.e., where daytime transects might have been reallo- 
cated and what the sardine densities may have been in 
those locations), it is not possible to quantify the likely 
reduction in sampling variance when the predicted 
sardine habitat is used to optimize sampling. 
When acoustic-trawl surveys of Pacific sardine are 
performed during spring, the sampling should extend 
to the southern limit of the potential sardine habitat, 
perhaps south of the U.S. -Mexican border (Felix-Uraga 
et al., 2004). When acoustic-trawl surveys of Pacific 
sardine are performed during summer, attention should 
be paid to the northern limit of the potential sardine 
habitat, perhaps north of the U.S. -Canadian border 
(Ware, 1999). Also, because sardine reside closer to the 
shore during summer, the eastern ends of the transects 
should be extended as close to shore as practical. 
Sampling optimization can involve considerations of 
many resource factors and scientific objectives, e.g., 
availability of ship time and single-species assessments 
versus ecosystem-based assessments of multiple spe- 
cies. Nevertheless, optimization of surveys of single or 
multiple species, or ecosystems will benefit from a priori 
characterization of the probable population boundaries. 
Unless a survey region is defined as larger than any 
possible extent of the target population! s), which may 
be cost prohibitive or increase sampling variance, pre- 
dictions of the population habitat! s) can be useful for 
optimally applying survey resources. On the other hand, 
if the survey region is defined somewhat arbitrarily 
and held constant, the proportion of the target stock(s) 
residing inside the survey area may vary between sur- 
veys, causing insidious variable hias(es). Although the 
accuracy of the habitat-model predictions may vary over 
time, they can be re-evaluated with the results from 
each survey, e.g., by quantifying the proportion of the 
target-species biomass residing near the edges of the 
modeled habitat. 
Conclusion 
Acoustic-trawl surveys can provide high-resolution, 
accurate and precise information about the biomasses, 
and geographic and vertical distributions of multiple 
species and trophic levels. Acoustic-trawl methods can 
be employed concurrently with other survey methods, 
perhaps enabling investigations of animal interactions 
with each other and their environments. 
The principal challenges of acoustic— trawl surveys 
of sardine are to estimate and survey the potential 
sardine habitat (habitat estimation); to identify the 
contribution of sardine backscatter to the total acoustic 
backscatter (species identification); and to estimate the 
mean acoustic backscatter per average-individual sar- 
dine (TS estimation). Reduced uncertainty in the survey 
results will likely result from a more optimal allocation 
of sampling effort in space and time, increased net sam- 
pling, and refinements to acoustic-target identification 
and TS estimation. 
Observed variations in time series of estimated abun- 
dance and distribution can result from population dy- 
namics or from random and temporally and spatially 
varying systematic errors. Therefore, to optimally apply 
survey effort and minimize both random and systematic 
errors in survey results, it is important to identify a 
priori the area likely to contain the stock, the potential 
sardine habitat, and allocate the available sampling 
effort to the area. This study shows that the model by 
Zwolinski et al. (2011) can be used for this purpose. 
Generally, the model indicates that spring surveys of 
the northern stock of sardine may be focused offshore of 
central and southern California and perhaps northern 
Baja California, and summer surveys may be focused 
