Lough and O'Brien: Life-stage recruitment models for Gadus morhua and Melanogrammus aeg/efinus on Georges Bank 
131 
20 30 
Age-1 recruits (no. x 10 6 ) 
Figure 6 
Demersal juvenile Atlantic cod ( Gadus morhua) mortality rate versus 
the previous year’s age-1 recruits on Georges Bank between 1977 and 
1987. The yearly demersal juvenile mortality rates were estimated 
by the exponential loss between the beginning of the demersal stage 
and the age-1 recruits. The density-dependent regression was used to 
adjust the demersal juvenile cod mortality rates in the time series. 
The same cod regression (Eq. 7) was used 
for haddock because cod are considered to be 
their main predator (Tsou and Collie, 2001). 
Demersal juvenile mortality rates derived 
by this method are listed in Tables 1 and 2. 
Results 
The yearly initial egg abundance derived 
from the MARMAP and GLOBEC ichthy- 
oplankton surveys were within the same 
order of magnitude as those derived from 
the VPA SSB fecundity relationships, but for 
the most part, the values did not track each 
other (Fig. 3). Estimates determined from the 
GLOBEC surveys for haddock had the best 
correspondence with the VPA estimates. Egg 
abundance for cod was higher than the VPA 
estimate in 9 of 14 years (Fig. 3A), whereas 
egg abundance determined for haddock was 
higher in only two years, 1980 and 1983 (Fig. 
3B). There was no significant correlation 
between the survey egg abundance and age-1 
recruits for either species; however, there 
was a significant correlation between age-1 
recruits and both larval and pelagic juvenile 
abundance (Tables 3 and 4). 
Correlations between recruitment of cod 
predicted with the various life-stage models and the 
cod VPA age-1 recruitment are shown in Table 5 and 
Figure 7. Starting from the spawning abundance time 
series, the best recruitment correlation was obtained 
from the GLOBEC years (r=0.94, P< 0.05) with the 
demersal juvenile mortality rate based on regression of 
the previous year’s recruitment, although both methods 
produced similar results. The recruitment values pre- 
dicted from the GLOBEC years were nearly identical to 
those predicted by the VPA; however, there was more 
disparity between the two values during the MARMAP 
years (Fig. 7A). The recruitment values predicted by the 
VPA were consistently somewhat higher than those pre- 
dicted from the spawning abundance model. The years 
1983 and 1987 did not track the pattern of these values 
estimated by the VPA as did the other years. Note there 
was no egg abundance data for 1977 and 1978. 
The hatching abundance models produced similar cor- 
relations with the recruitment values from the VPA as 
the spawning abundance models. The GLOBEC years 
had the best correlation (r=0.93, P<0.05) with nearly 
identical predictions (Fig. 7B). The MARMAP year’s re- 
cruitment values were lower than VPA values, especial- 
ly again for 1983 and 1987, and the two additional lar- 
val years, 1977 and 1978. The GLOBEC years provided 
a more reliable comparison than the MARMAP years 
since surveys were conducted monthly with more sam- 
pling stations than the bimonthly MARMAP surveys. 
Starting from the seasonal egg abundance derived 
from the VPA SSB and using the derived proxies for 
estimating egg and larval mortality estimates to predict 
Table 3 
Coefficients of correlation (r) among Atlantic cod ( Gadus 
morhua) early life-stage abundances determined from 
ichthyoplankton surveys, and recruitment at age-1 based 
on virtual population analysis abundances for 1977-87, 
and 1995-99. Life-stage abundances were log 10 trans- 
formed. Correlations where P< 0.05 are starred. 
Life stage 
Eggs 
Larvae 
Pelagic 
juveniles 
Recruits 
age-1 
Eggs 
1.00 
Larvae 
0.23 
1.00 
Pelagic juveniles 
0.15 
0.70* 
1.00 
Recruits age-1 
-0.20 
0.57* 
0.36 
1.00 
recruitment, we calculated the correlations with VPA 
recruitment values for the MARMAP and GLOBEC 
years, and for the full time series, 1977-2004 (Table 5). 
Egg mortality was based on the wind stress model. The 
demersal juvenile mortality rate based on regression of 
recruitment from the previous year was used because 
it generally produced a better correlation. The low cor- 
relation (r=0.37) for the full time series is evident in 
Figure 7C, poorer for the MARMAP years (r=0.25), and 
even poorer for the GLOBEC years (r=0.05). 
Correlations between the various predictions of re- 
cruitment values with the haddock life-stage models 
and the haddock VPA age-1 recruitment values are 
