Demer et at: Seasonal migration of Sardmops sagax in the California Current Ecosystem 
53 
months of 2009 and 2010 (Hill et al., 2010). The results 
were included in the assessments for those years, but 
their influence was marginal because the random sam- 
pling errors were large and the portions of the stock sur- 
veyed were uncertain (Hill et al., 2010). Consequently, 
the PFMC sought a fisheries-independent survey method 
that could accurately and precisely quantify the entire 
northern stock of sardine. 
Acoustic-trawl surveys 
Combining information collected with sonars and nets, 
“acoustic-trawl” methods (those combining echosounder 
and trawl sampling) have been used to survey sardine off 
the west coast of the United States, within the CCE, for 
more than a half century. Beginning with “sonar map- 
ping” in the 1950s (Smith, 1978), and single-frequency 
echo-sounding in the 1960s (Mais, 1977), the acoustic 
survey equipment and methods evolved to “broad band- 
width resonance scattering” in the 1970s (Holliday, 1972; 
1977), and now to multiple-frequency and multibeam 
echo-sounding (e.g., Cutter Jr. and Demer, 2008). In all 
cases, interpretation of the acoustic backscatter data 
relies on information, including fish species and their 
distributions and sizes, from periodic net catches. 
Sardine habitat and migration 
Potential habitat To minimize uncertainties in esti- 
mates of sardine biomass, irrespective of the survey 
technique, the sampling effort must be optimally allo- 
cated to only the region containing the stock (Simmonds 
and MacLennan, 2005). On the basis of water tempera- 
tures associated with spawning activity and on evidence 
that sardine may be food limited above 16.5 °C, potential 
spawning-sardine habitat has been described as seawa- 
ter with temperatures from 14° to 16°C (Jacobson et al. 1 ), 
13.5° to 16.5°C (Agostini, 2005), and 12° to 15°C (Reiss 
et al., 2008). Notwithstanding these observed associa- 
tions, accurate predictions of sardine distributions and 
densities have been elusive, until recently. Zwolinski et 
al. (2011) demonstrated accurate predictions of potential 
sardine habitat and the dynamics of its spatiotemporal 
distribution. 
Based on a 12-year data set including samples of 
sardine eggs and concomitant remotely sensed oceano- 
graphic conditions, a probabilistic, generalized-additive 
model was developed which predicts the distributions 
of habitat for the northern stock of sardine. Significant 
relationships were identified between sardine eggs and 
sea-surface temperature, chlorophyll-a concentration, 
and the gradient of the sea-surface height. The model 
describes and accurately predicts the potential habitat 
and seasonal migration pattern of sardine, whether or 
not they are spawning (Fig. 1). The model predictions of 
1 Jacobson, L. D., N. C. H. Lo, S. F. Herrick Jr., and T. Bishop. 
1995. Spawning biomass of the northern anchovy in 1995 
and status of the coastal pelagic species fishery during 
1994. NOAA Admin. Rep. LJ-95-11, La Jolla, 49 p. 
potential habitat were extensively validated by fishery- 
landing data from Oregon, Washington, and British 
Columbia, and trawl-survey data collected near the 
Columbia River mouth. 
The predicted habitat can be used to optimize the 
locations for sardine surveys. For example, Zwolinski 
et al. (2011) showed that, averaged over twelve spring 
surveys, 92% of the sardine eggs were sampled by us- 
ing 64% of the original survey effort. In other words, 
habitat predictions could have allowed approximately 
36% of the survey effort to be reallocated to potential 
habitat — thus likely reducing the sampling error. The 
model of potential sardine habitat can also be used to 
optimize the survey timing. 
Seasonal migration Traditionally, DEPM surveys of 
CPS have been performed in the spring, during the 
peak of the sardine spawning season (Lo et al., 2009). 
At that time, CPS mostly aggregate offshore of central 
and southern California, but some species, particularly 
Pacific herring ( Clupea pallasii) and northern anchovy 
( Engraulis tnot'dax), are located in a few coastal areas 
farther north. However, the model of potential sardine 
habitat indicates that acoustic-trawl surveys of sardine 
may be most efficiently conducted during the months of 
June and July, when the habitat is compressed along the 
coasts of Oregon and Washington (Fig. 1), the fish are 
located generally north of Point Conception and south 
of the Strait of Juan de Fuca, the days are longest and 
thus daytime sampling is maximized, and the survey 
can be augmented with fishery catch data from the same 
general time and place (Zwolinski et al., 2011). 
In this study, model predictions of potential sardine 
habitat (Zwolinski et al., 2010) are compared to the 
results of two acoustic-trawl surveys of the entire west 
coast of the U.S., in spring and summer 2008. The 
principal objectives are 1) to estimate the geographic 
distributions and biomasses of the northern stock of 
sardine during the southern and northern portions of 
their migration; 2) to estimate random and systematic 
sampling errors in those estimates; 3) to further vali- 
date the model predictions of potential sardine habitat 
and its seasonal dynamics; and 4) to evaluate the fea- 
sibility of simultaneously surveying other CPS. 
Materials and methods 
Survey design 
The west coast of the United States (U.S.) was surveyed 
twice in 2008, during spring (25 March to 30 April) and 
summer (6 July to 18 August). Sampling during the 
spring survey was conducted from the NOAA research 
vessels (RV) David Starr Jordan and Miller Freeman. 
Sampling during the summer survey was conducted from 
RV David Starr Jordan. During both seasons, the survey 
tracks (Fig. 2) mostly spanned and often extended well 
beyond the potential habitat of sardine (Fig. 1). A large 
amount of survey effort was expended outside of the 
