446 



Fishery Bulletin 90(3). 1992 



40 

 Age (Days) 



Figure 6 



Effects of altering length of the early-larval stage on 

 recruitment correlation for the cod DI-COV model. 

 Early stage is defined as having twice the daily mor- 

 tality rate of the total larval period. Shown are results 

 with the early stage set at 5 (■). 10 (♦). and 15 (D) 

 days. Symbols represent, from left to right: egg, yolk- 

 sac larvae, early larvae, and metamorph stages. 



Discussion 



My results indicate that only predictions of recruitment 

 based on abundances of postmetamorphic fish are likely 

 to be useful for the management of marine fishes. The 

 contribution to recruitment variation made by egg 

 number (and, therefore, stock biomass) is very small, 

 a prediction confirmed by most stock-recruit data (Par- 

 rish 1973). Correlations involving the abundance of 

 early larvae are stronger, but are still too weak for 

 forecasting. The model R'-^ values for correlations in- 

 volving early larvae are similar to the range, extending 

 from 0.01 to 0.66, for published values compiled by 

 Peterman et al. (1988). Accurate recruitment fore- 

 casting may be possible by sampling during the late- 

 larval period (Graham and Sherman 1987). However, 

 this is highly dependent on parameters and the dy- 

 namics of the particular species; only for cases with low 

 variability in juvenile mortality or with mortality rates 

 correlated across stages are the abundances of late lar- 

 vae likely to be useful for recruitment forecasting. 



Research on recruitment variability has been ori- 

 ented to the early-larval stages largely as the result 

 of Hjort's (1913) hypotheses and the observation that 

 most of the individuals of a year-class die during the 

 first few weeks of life (Wooster and Bailey 1989). In 

 my four example species, the average cumulative mor- 

 tality on the cohort to the 10-d larval stage is 93% 

 (Table 1), yet the variation in abundance of these lar- 

 vae explains more than 50% of recruitment variability 



in less than half of the cases. Variability in the late- 

 larval and juvenile stages is still large enough to in- 

 fluence the strength of correlations of recruitment with 

 larval abundances. My results suggest that the stage 

 'when year-class strength is determined' (defined here 

 as i?^>0.5) occurs after this early critical period. 

 However, the sensitivity analysis indicates that the 

 strength of the correlation between the abundance of 

 larvae and recruitment will depend strongly on the rate 

 at which mortality declines during the larval period, 

 and at what age the larvae are being sampled. 



Strong linkages in mortality rates across intervals 

 also render the definition of a 'critical period' less con- 

 cise. Correlations between early life stages and recruit- 

 ment were stronger when there were linkages, because 

 survival to the age of sampling will be correlated to 

 survival in the future. An estimate of mortality or abun- 

 dance in one stage will be an index of mortality in all 

 early life stages. This may be especially true for the 

 early-larval stages (the classical 'critical period') be- 

 cause small larvae are probably subjected to a similar 

 source of mortality as older larvae, especially if spawn- 

 ing occurs over a protracted period, mixing larvae of 

 different ages together in the same body of water. In 

 addition, environmental conditions during an early 

 stage may affect survival of the cohort in the future. 

 Poor feeding conditions of early larvae, for example, 

 may have a long-term effect on growth and survival 

 (Frank and McRuer 1989). In these cases, recruitment 

 will be somewhat predictable from the early-larval 

 stages, but this is not support for a strict interpreta- 

 tion of Hjort's hypothesis that an early critical period 

 determines recruitment because mortality is correlated 

 across all prerecruit stages. 



The difficulty and expense of obtaining accurate 

 estimates of abundances of eggs and larval fish have 

 led to increased interest in finding indirect estimates 

 of mortality rates that may be simpler to collect and 

 could provide an index of year-class strength. Such 

 measures include estimates of growth (Houde 1987), 

 condition, lipid content (Theilacker 1986), and RNA/ 

 DNA ratios (Buckley and Lough 1987) as well as ocean- 

 ographic variables such as upwelling and wind events 

 (Peterman and Bradford 1987). My results show that 

 such mortality estimates made on small larvae are not 

 likely to be strongly correlated with recruitment 

 (Fig. 4). Mortality estimates on older larvae will have 

 stronger correlations, potentially closer to a value of 

 0.50. Note that the correlations in Figure 4 are for 

 direct estimates of mortality, indirect indices will be 

 more poorly related to recruitment. A combination of 

 larval abundances and mortality rate estimates may 

 allow more precise prediction of recruitment (Graham 

 and Sherman 1987, Frank and McRuer 1989); if esti- 

 mates are accurate and are based on older larvae, 



