486 



Fishery Bulletin 97(3), 1999 



10^ 



7.5 



01 



< 



5- 



2.5 



the first two weeks of April had no 

 samples at all for the second half of 

 the month. To make maximum use of 

 the data available, estimates of the 

 percentage of egg-bearing shrimp from 

 individual samples were averaged for 

 the first and second halves of the 

 month, respectively. These semi- 

 monthly figures were then averaged 

 to produce an overall mean percent- 

 age of egg-bearing shrimp in the catch 

 for each April. When samples were miss- 

 ing from the latter half of the month, a 

 zero level of ovigerous shrimp was as- 

 sumed. When samples were missing for 

 the first half of the month, the level of 

 ovigerous shrimp observed for the sec- 

 ond half was assumed as a minimum 

 estimate for the first half of April. Ac- 

 cordingly, the estimates presented in the 

 present study are minimum estimates 

 of the catch of egg-bearing shrimp. 



To determine when, during the 

 month of April, egg-bearing shrimp had declined to 

 a minimal component of the catch, I constructed a 

 scatter graph of the raw percentages of egg-bearing 

 females by date, using existing data for 1961-97 

 (ODFWM. Area 82 (Fig. 1) was excluded from this 

 analysis because the incidence of egg-bearing shrimp 

 in this area was generally very low in April. 



A stepwise process was used to evaluate the rela- 

 tionships between the new recruitment, spawner, and 

 egg production indices, calculated at assumed val- 

 ues of 0.25., 0.50 and 0.75 for trawl efficiency. Fol- 

 lowing the findings of my earlier study (Hannah, 

 1993), the first step in this analysis was to regress 

 log (natural log unless noted) recruitment against 

 April SLH at Crescent City, California, to verify that 

 SLH still explained a significant amount of the 

 interannual variation in recruitment. Simple linear 

 regression was used for this test. These SLH data 

 were obtained from the National Oceanic and Atmo- 

 spheric Administration's Ocean and Lake Levels Di- 

 vision for the years 1979-95, corresponding to age-1 

 catch years 1980-96. Next, the residuals were ex- 

 amined and compared graphically and by means of 

 linear regression to the spawner and egg production 

 indices, and outliers were examined in light of the 

 estimates of the catch of egg-bearing shrimp in April. 

 Finally, some multivariate models incorporating 

 April SLH and the spawner and egg production indi- 

 ces were fitted by using multiple regression. 



Garcia ( 1983) and others have stressed that many 

 of the stock-recruitment relationships that have been 

 demonstrated for shrimp stocks are really statisti- 



80 



60 



40 



-20 



1980 



1985 



1990 



1995 



Year 



Figure 2 



Time series of the recruitment index, calculated assuming elemental trawl 

 efficiency of 0.50. and the fishery catch of egg-bearing shrimp from April of 

 the year of larval release, by year of age-1 recruitment, 1980-96. 



cal artifacts caused by a strong recruitment-stock 

 relationship (the reverse of a stock-recruitment re- 

 lationship, rather spawning stock in year / as a func- 

 tion of recruitment that same year) in combination 

 with serially autocorrelated, environmentally driven, 

 recruitment. To test for such a possibility in the 

 present analysis of ocean shrimp recruitment, I re- 

 gressed the new recruitment index against itself at 

 a lag of one year. I also examined the recruitment- 

 stock relationship further by graphically comparing 

 the spawning index, broken out by age class, to the 

 recruitment index from the same year. 



Results 



The recruitment index (Fig. 2; Table 1; all indices 

 calculated assuming 0.50 for trawl efficiency, unless 

 noted) shows wide variation in ocean shrimp recruit- 

 ment, especially after 1986 (year of age-1 recruit- 

 ment, unless noted). Recruitment in 1983, 1984, 

 1990, and 1993 was very low, whereas very high re- 

 cruitment was observed for the 1987-89 and 1992 

 year classes. These four large year classes had a very 

 significant influence on total landings from the 

 shrimp fishery during the years included in this 

 study. Although they represent less than one third 

 of the years in the time series, they contributed 

 roughly 60% of the total landings of shrimp. 



The log of the shrimp recruitment index was 

 stongly negatively correlated with April SLH at Cres- 

 cent City, California, in the year of larval release 



