444 



Fishery Bulletin 90(3). 1992 



early larvae and recruitment than the separate larval 

 regression. 



Correlations between 



early life history and recruitment 



Correlations between abundances at early life stages 

 and recruitment increased in strength as the interval 

 between the two stages decreased (Figs. 2, 3). Overall, 

 covariances in mortality rates across stages increased 

 i?2 values between early abundances and recruitment 

 by 0.01-0.25, while density-dependent juvenile mortal- 

 ity had only a small and usually negative effect onR^ 

 values. 



Correlations between egg or first-feeding larvae and 

 recruitment were weak; the average R^ over all spe- 

 cies and models was 0.05 for eggs and 0.20 for first- 

 feeding larvae. None of the values exceeded 0.50, in- 

 dicating that these early stages have little predictive 



as 



CO 



400 



1 

 0.8 

 0.6 

 0.4 

 0.2 



100 200 



Age (Days) 



300 



400 



Figure 2 



Predicted R~ values for correlations between recruitment 

 and early life stages for cod (O), anchovy (■), plaice (»), and 

 herring (D). For each species, symbols represent, from left 

 to right: abundance of eggs (at t = 0), first-feeding larvae, 10-d 

 larvae, and metaniorphs. Dotted line indicates the strength 

 of correlations required for recruitment prediction (Walters 

 1989). Both examples include density-independent juvenile 

 mortality; lower panel also incorporates covariance between 

 interval-specific mortality rates. 



capability. At the end of the early-larval stage, R^ 

 values increased; and in 4 of 16 cases in Figures 2 and 

 3 the R~ values exceeded 0.50. However, no values 

 exceeded 0.80, the suggested requirement for recruit- 

 ment forecasting to be beneficial for management 

 (Walters 1989). 



In nearly all cases, the majority {R^>0.50) of re- 

 cruitment variation was predictable at the age of meta- 

 morphosis. The exception was the herring example, 

 which gave low correlations because of high variabil- 

 ity in the juvenile mortality rate. Half the correlations 

 met the forecasting requirement of i?">0.80 by the 

 age of metamorphosis; these cases occurred in species 

 with the lowest juvenile mortality rates. 



The success of larval mortality rates in predicting 

 recruitment was lower than for larval abundance esti- 

 mates. The correlation between the mortality rate of 

 the early-larval period and recruitment was strongly 

 affected by the presence of covariation between stage- 

 specific mortality rates; without these covariances the 

 average R- was 0.12; the largest value was 0.18. 

 When the covariances were incorporated, these correla- 

 tions are increased, although none exceed 0.5 (Fig. 4). 

 Ten of 16 R- values exceeded 0.50 for the much 



400 



Density Dependent 

 Covariance 



1 00 200 300 



Age (Days) 



400 



Figure 3 



As in Figure 2, except both versions include density-dependent 

 juvenile mortality; lower panel also incorporates mortality 

 covariances. 



