64 
Fishery Bulletin 110(1) 
advance of its arrival (Ona et al., 2007). Also, the spa- 
tial distributions of acoustically detected CPS matched 
well those of the trawl catches in areas of high biomass. 
Furthermore, despite the sardine residing offshore dur- 
ing spring and shallower near the coast in the summer, 
the two 2008 acoustic-trawl survey estimates of sardine 
biomass were not statistically different from each other 
or from the assessment estimates. In other words, if 
sardine avoid a vessel significantly, it is likely that the 
associated bias would increase when the fish naturally 
reside in shallower water; however, there is no evidence 
of this effect. 
Species identification and TS estimation 
The echo energy was apportioned to species by using 
a numerical algorithm that incorporates the following 
assumptions: 1) echoes from fish schooling in the upper 
70 m during the day can be identified as CPS by their 
backscattering spectra; 2) a representative portion of 
those CPS are sampled with the surface trawl at night; 
and 3) the TS-weighted proportions of the various CPS 
in the catches can be used to apportion the nearby CPS 
echoes to species. Because the distributions of the CPS 
echoes matched those of the CPS caught in the trawl, 
these assumptions appear to be valid. Where CPS were 
acoustically mapped, they were caught in the trawls; 
where CPS were not acoustically observed, they were 
absent from the catches (Fig. 2), in general. Further- 
more, the distributions of catches showed some degree of 
segregation among the various species, which supports 
the method of ascribing CPS backscatter to species based 
on their proportions in the nearest catches. 
Fish behaviour can affect trawl sampling. If certain 
species or sizes avoid capture, “net selectivity” causes a 
variable sampling bias. With the acoustic— trawl method, 
it is currently assumed that the net sampling is unbi- 
ased and therefore the proportions of CPS in the catch, 
and their length distributions, are representative of 
their respective stocks. However, there may be some net 
selectivity which will affect the species identifications 
and TS estimations, and cause variable sampling biases 
in the biomass estimates. 
In the absence of TS models for the target species 
in the populations and conditions under study, the bio- 
mass estimates were computed by using TS-to-biomass 
relationships derived for related species in similar sys- 
tems (Barange et al., 1996). The TS of fish with swim 
bladders are intrinsically variable, depending mainly 
on the acoustic frequency and the swim bladder size 
and orientation relative to the incident sound wave 
(Foote, 1980). The swim bladder size and orientation 
are related to fish anatomy, physiology, behavior, and 
ontogeny (Ona, 1990). Consequently, the TS-to-biomass 
relationships should ultimately be derived from mea- 
surements of target fish in the conditions under which 
they are sampled (Fassler et al., 2008). Future studies 
should evaluate uncertainty in the TS models; and new 
functions should be tailored for the populations in the 
CCE, accounting for acoustic frequency, fish length and 
depth, and season. For example, high-resolution images 
from X-rays (e.g., Conti and Demer, 2003; Renfree et 
al., 2009) or magnetic resonance (e.g., Pena and Foote, 
2008) can be used to parameterize scattering models 
and better predict TS as a function of acoustic fre- 
quency, and fish morphometries, depth, and orientation 
(e.g., Horne, 2003; Cutter Jr. and Demer, 2007; Cut- 
ter Jr. et al., 2009). The frequency response of single- 
and mixed-species aggregations can then be simulated 
by summing the responses of fish varying in number, 
depth, and orientation. 
Seasonal migration 
Sardine were distributed in the south and offshore 
of southern and central California in the spring and 
were compressed along the coast, mainly from northern 
California to Washington, in the summer (Fig. 3). These 
findings are consistent with the predictions of seasonal 
changes in potential sardine habitat in the CCE (Fig. 
1; Zwolinski et al., 2011). Small discrepancies, par- 
ticularly in the dynamic nearshore upwelling areas, 
can be attributed to the density-dependent nature of 
sardine habitat use and temporal mismatches between 
the oceanographic conditions during the shipboard 
sampling and the multiweek averages used to map the 
habitat. 
The potential sardine habitat annually oscillates be- 
tween north and south as a consequence of seasonal 
oceanographic changes. On the basis of the similari- 
ties of the spring and summer estimates of sardine 
biomass, seasonal migration appears to have involved 
the entire population. During the summer, few sardine 
were mapped off central and southern California, south 
of 40°N (Fig. 3), suggesting that the inflow of individu- 
als from the southern stock (Felix-Uraga et al., 2005) 
was negligible in summer 2008, most likely because the 
CCE was colder than average owing to La Nina condi- 
tions (McClatchie et al., 2009). 
Jack mackerel also appear to be affected by the same 
mesoscale forcing to which sardine are subject. They 
are recurrently mapped in the warmer margins of the 
potential sardine habitat (Fig. 3). However, the north- 
ward migration of jack mackerel during the summer 
was not as marked as that of sardine. Pacific mackerel 
were scattered in offshore and coastal waters — usually 
among the more abundant jack mackerel and sardine, 
probably schooling with them. 
In contrast to the predicted and observed migrations 
described above, northern anchovy exhibit a stronger 
geographic fidelity and were found in expected discrete 
locations in the southern California Bight (Smith and 
Hewitt, 1985), and off Oregon and Washington (Laroche 
and Richardson, 1980). 
Future surveys 
Although the acoustic-trawl method can be further 
refined, the results appear to be precise and accurate. 
The CV values are low, and the spring and summer 
