Weber and McClatchie: Effect of environmental conditions on the distribution of Scomber japonicus 
95 
mixture of geographically offset but overlapping distri- 
butions. A drawback of this approach is that the actual 
geographic centers and relative proportions of the two 
spawner types likely change through time, and these 
changes are included in the unexplained variance of 
the model. However, given the large overlap in spawn- 
ing behaviors (Fig. 3, D and E) and lack of ability to 
discriminate between the two spawner types, this may 
be the only practical way to evaluate habitat effects 
on their distributions. Latitude also partly reflected 
the distribution of habitat from north to south. For 
example, salinity has frequently been used as an in- 
dicator of water masses preferred by pelagic fish (e.g., 
Checkley et al., 2000). Salinity was not used in this 
model because it had a relatively high correlation with 
latitude (r= 0.58). 
The core area consistently had greater zooplankton 
displacement volumes but less appropriate (slower) geo- 
strophic flows for Pacific mackerel larvae than Mexican 
waters (Fig. 5, B and C). The most favorable average 
annual temperatures (Fig. 5A) alternated between the 
core area and Mexican waters. Mexican waters more 
fully represented the transition zone occupied by both 
types of spawners: those that were predicted to pre- 
fer spawning in April in 15.5°C water, and those pre- 
dicted to prefer spawning in August in about 22°C 
water. In contrast, when temperatures departed from 
the optimum in the core area, usually they were too 
cold for either group, resulting in better habitat condi- 
tions in Mexican waters (e.g., the early to mid 1950s). 
The greater zooplankton displacement in the core area 
suggests that greater productivity may have attracted 
some Pacific mackerel to the SCB that otherwise pre- 
ferred warmer temperatures. This idea is consistent 
with data from a tagging study that indicated Pacific 
mackerel migrated between the two areas seasonally, 
moving northward in summer and southward in winter 
(Fry and Roedel, 1949), and that the migration was 
more pronounced during El Nino events. For Pacific 
mackerel that prefer warmer water, movement between 
Mexican waters and the SCB likely represents a trade- 
off between optimal temperatures and greater feeding 
opportunities. 
The precise mechanism underlying the relation be- 
tween geostrophic flow and Pacific mackerel larvae in 
this study is unknown. Geostrophic flow was included 
as a predictor because it was hypothesized to be related 
to the productivity field (Mantyla et al., 2008) and has 
been previously related to abundance of Pacific mack- 
erel (Yatsu et ah, 2005). However, geostrophic flow is 
related to productivity through adjustment of the water 
column in response to the current, yet no direct relation 
between larvae and mixed-layer depth was measured 
in this study. Either the field of geostrophic flows cal- 
culated for each year in this study provided a better 
measure than the calculated mixed-layer depths at each 
CalCOFI station, or geostrophic flow affected larval 
distribution through another unknown mechanism. 
The CPFV index for the previous year had a larger 
effect on model predictions than any of the variables re- 
lated to habitat (Fig. 5, F-H). After correcting for stock 
size and sampling time, mean annual differences in 
predicted probability of capture varied by less than 7%, 
indicating that habitat quality was much more stable 
among years than was stock size. The importance of the 
CPFV index indicates that Pacific mackerel larvae did 
not fully occupy the suitable habitat during most years. 
Some of the best habitat for larvae was predicted to oc- 
cur near Punta Eugenia in the early 1950s, but catches 
were small, in part, because the stock size was small 
(cf. Fig. 2 and Fig. 6). Likewise, recent low catches in 
the 2000s appear related to small stock size and po- 
tentially other unknown factors, but the environmental 
conditions modeled in this study have remained almost 
as suitable for Pacific mackerel larvae as they were in 
the 1990s. These results are consistent with previous 
studies that indicated that the stock-recruit relation- 
ship for Pacific mackerel in the Northeast Pacific is 
not strong (Parrish, 1974; Parrish and MacCall, 1978). 
The relatively small changes in quality of larval habitat 
predicted by the model are unlikely to have large effects 
on future recruitment success or stock size. 
Predicted probabilities of capture varied between the 
core area and Mexican waters when both areas were 
sampled (Fig. 4), particularly after correcting for day of 
sampling and stock size (Fig. 6). These results suggest 
analyses that rely on data from the core area alone as 
an index of the entire population likely contain some bi- 
as. A model such as the one reported here could be used 
to tune a time series of larval production in the core 
area by scaling years up or down according to mean 
habitat conditions. So, for example, larval production 
would be assumed to be greater than that measured in 
the SCB during cold years, when a larger proportion of 
the stock presumably spawns in Mexican waters. How- 
ever, we do not recommend such an approach given the 
statistical variability associated with this type of model. 
It would be much better to include data collected in 
Mexican waters as part of the IMECOCAL program in 
future studies and assessments. Although IMECOCAL 
data were not consistently available for previous LT.S. 
stock assessments (Crone et al., 2009), recent analyses 
have been conducted with integrated data from both 
programs (e.g., Lo et al., 2010). We suggest that further 
analyses with integrated data sets would allow both 
nations to achieve better assessments with less bias. 
The interaction between day of year and latitude was 
an important predictor in the model, indicating that 
some samples were more likely to contain larvae than 
others simply because sampling was conducted when 
Pacific mackerel were more likely to be spawning at 
the sample location (Fig. 5D). This problem would also 
create some bias in estimates of larval production be- 
cause larval production estimates (Lo et al., 2010) do 
not contain a correction for the fraction of the adults 
spawning when sampling occurs (unlike the daily egg 
production method, cf. Lasker, 1985). In practice, the 
bias is likely to be small for annual estimates of Pa- 
cific mackerel production in the SCB because CalCOFI 
cruises occur in April and July, near the beginning and 
