AbStTclCt. A simple analytical 



technique is developed for estimating 

 the predictability of recruitment, that 

 is, correlations between recruitment 

 and stage-specific mortalities or abun- 

 dances. The method requires the input 

 of estimates of the variability of stage- 

 specific mortalities, which may be cal- 

 culated from mean stage-specific mor- 

 talities by applying a published regres- 

 sion. It is shown that modification of 

 this regression to compensate for sam- 

 pling error in field measurements of 

 abundance significantly reduces the 

 estimated standard deviation of log-re- 

 cruitment, which is an important fac- 

 tor in the predictability calculations. It 

 is concluded that the prospects for pre- 

 dicting recruitment from egg or larval 

 surveys or from environmental vari- 

 ables are quite poor for fish stocks 

 showing the typical distribution of mor- 

 tality across stages. 



Estimating the predictability of 

 recruitment 



Gordon Mertz 

 Ransom A. Myers 



Science Branch, Department of Fisheries and Oceans 



Northwest Atlantic Fisheries Centre 



PO. Box 5667, St. John's, Newfoundland A1C 5X1 



Manuscript accepted 7 May 1995. 

 Fishery Bulletin 93:657-665 ( 1995). 



The problem of predicting recruit- 

 ment remains central to fisheries 

 science (e.g. Bradford, 1992). Ap- 

 proaches to this task may involve 

 finding environmental correlates of 

 recruitment or the field sampling of 

 prerecruit life history stages. In this 

 study we present simple analytical 

 formulae that permit one to esti- 

 mate the potential explainable vari- 

 ance of recruitment without the use 

 of detailed, specific data. 



Certain environmental factors 

 may be correlated with recruitment. 

 Wind speed has been proposed as a 

 determinant of recruitment because 

 storm-driven mixing can disperse 

 larvae and their prey reducing food 

 availability (Lasker, 1975, 1981; 

 Buckley and Lough, 1987; Peter- 

 man and Bradford, 1987). Larval 

 food supply may also be influenced 

 by the lag between appearance of 

 larvae and the peak abundance of 

 their prey (Cushing, 1990). The in- 

 tensity of turbulence may control 

 the frequency of contact between 

 larvae and their prey (Rothschild 

 and Osborn, 1988). Larvae may be 

 exported to inhospitable waters by 

 the action of wind driven currents 

 (Nelson et al., 1977) or by the in- 

 cursion of Gulf Stream rings (Flierl 

 and Wroblewski, 1985; Myers and 

 Drinkwater, 1989). (For thorough 

 discussions of environmental influ- 

 ences on recruitment see Fogarty 

 [1993] or Wooster and Bailey 

 [1989].) In each example noted 



above, some measurable physical 

 quantity may be plausibly postu- 

 lated to be a proxy for (say) larval 

 mortality, in a qualitative sense; it 

 is our aim to quantify the expected 

 predictive power of an environmen- 

 tal variable. Alternatively but much 

 more expensively, larval mortality 

 could be estimated from field stud- 

 ies (Butler, 1991). We calculate the 

 likely strengths of the correlations 

 between mortality for an early life 

 history stage and recruitment. A 

 related problem that is addressed 

 is the correlation between recruit- 

 ment and abundance in an early life 

 history stage, which can be deter- 

 mined from field studies (Pete rman 

 et al., 1988; Bradford, 1992). This 

 treatment is an analytical comple- 

 ment to the simulation studies pre- 

 sented in Bradford (1992). 



In the analysis to follow we first 

 show how variability of mortality 

 may be estimated from mean mor- 

 tality while accounting for the ef- 

 fect of sampling error in the field 

 measurements. We then proceed to 

 formulate simple relationships per- 

 mitting the calculation of the cor- 

 relation coefficients between log- 

 transformed or raw recruitment 

 and stage-specific mortality or 

 abundance, using only estimates of 

 variability in stage-specific mortal- 

 ity. These two sets of analyses are 

 then combined to provide estimates 

 of the predictability of recruitment 

 for a number of fish species. 



657 



