Results
As judged by the BIC, and confirmed by the pseudo-R2,
the mean length of copepods was more strongly related to SST than it was
to Latitude or Oxygen concentration (Table 1). We found that in all
cases, SST led to an improved fit (log-likelihood ratio, Table 1), so
SST instead of Latitude or Oxygen concentration was used in all
subsequent models.
The final model for mean copepod length included SST, Chl-a ,
Predation, and Proportion of Omnivores as predictors, and random effects
for Survey, and Tow within. All predictors reduced BIC and increased the
pseudo R-squared (Table 1). Together, fixed effects explained 64.2% of
the variance in mean length of copepods (Table 1). Random effects
explained a further 13.5%, where 1.8% was attributed to Survey (Fig.
2a), and 11.7% to the Tow within Survey effects (Fig. 2b).
The mean length of copepods declined with warmer SST (Fig. 2c), where
the mean length of copepods decreased by a factor of 0.982 for each 1ºC
increase in SST, which is equivalent to a 0.83 mm decrease across the
temperature range (-1.7–30ºC) or a 44.7% decrease in mean length. This
equates to an approximate linear decrease in copepod mass of
~2.6% per ºC, equivalent to an 83.0% decrease in mass
across the entire temperature range. Using a linear approximation, we
found for each 1ºC increase in temperature, the mean length of copepods
was ~0.026 mm shorter.
The mean length of copepods also declined with increased Chl-a(Fig. 2d), where for each square increase in Chl-a (based on
square-root transformation), the mean length of copepods decreases by a
factor of 0.811. This is equivalent to a 0.78 mm decrease across the
entire Chl-a range (0.02–9.51 mg.m-3) or a
46.0% decrease in mean length. This equates to an 84.1% decrease in
copepod mass across the Chl-a range. Using a linear
approximation, we estimate for each square increase in Chl-a that
copepods are ~0.265 mm shorter.
Further, the mean length of copepods increased in samples with higher
abundances of invertebrate predators (Fig. 2e), where for each e-fold
increase in predator abundance (an increase by a factor of 2.718), the
mean length increases by a factor of 1.011. This is equivalent to a 0.17
mm increase across the entire predation range, or an 11.3% increase in
mean length. This equates to a 37.7% increase in copepod mass across
the predation range. Using a linear approximation, we estimate for each
e-fold increase in predator abundance, copepods were
~0.018 mm longer.
The mean length of copepods also decreased in samples that had a higher
proportion of omnivorous copepods (Fig. 2f). The mean length was 2.34 mm
or 61.6% smaller in samples comprising only omnivores when compared
with samples comprising only carnivores. This equates to a 94.3%
decrease in copepod mass.
Investigating the relative importance of fixed effects showed that both
SST and Chl-a were of similar importance in determining the mean
length of copepods, as indicated by their BIC and
pseudo-R2. Removing either of these substantially
degraded BIC and pseudo-R2 (Table 1). Removing
Predation from the model resulted in a smaller decline in BIC and
pseudo-R2 (Table 1) and is thus less important in
determining mean length than SST and Chl-a . Finally, removing the
proportion of omnivores within samples degraded the BIC and
pseudo-R2 the most (Table 1).