320 



Fishery Bulletin 91(2), 1993 



lation between abundance of larvae and of later early 

 stages or between larval abundance and subsequent 

 recruitment. Because our explanatory data do not in- 

 clude wind speed, we were unable to address this spe- 

 cific question. Cury & Roy (1989) demonstrated that 

 spawning success in coastal pelagic fishes responds 

 nonlinearly to upwelling, but our linearized model could 

 not include this possible relationship. 



Methot (1989) has shown that the age of maturity in 

 northern anchovy increases from 1 to 2 when tempera- 

 tures are below average. Our data already account for 

 this effect, because the spawning-stock estimates of 

 Methot ( 1989) incorporate this effect of temperature. 



The only other likely determinant of anchovy re- 

 cruitment is spawner abundance itself. This is a 

 straightforward phenomenon in which very low 

 abundance produces very low recruitment. Our 

 stock-recruitment plot (Fig. lc) does not illustrate this 

 phenomenon; however, the plot begins after the lower 

 abundances of the 1950s, when the effect would have 

 been most pronounced. 



Pacific sardine 



As in the northern anchovy, the spawning biomass of 

 Pacific sardine depends strongly upon immediately- 

 preceding recruitments, with peaks in recruitment pre- 

 ceding peaks in spawning biomass by about 2yr 

 (Fig. 2a). Interpretation of the stock-recruitment plot 

 (Fig. 2c) is problematic: If the 2yr of highest spawn- 

 ing biomass (1949 and 1950) are disregarded, the plot 

 appears biphasic, with small spawning biomasses pro- 

 ducing small recruitments and large biomasses pro- 

 ducing large recruitments. But if all years are included, 

 recruitment appears to be density-dependent (or 

 strongly influenced by external factors). The stock-size 

 effect estimated by the model with climate effects was 

 not significant (P=0.66); we refit the model without a 

 stock-size effect (Table 5), reducing the R 2 by only 17c. 

 The model with combined climate and contaminant 

 effects identified a statistically significant stock-size 

 effect with the correct sign for compensation (Table 5). 

 This model also fit the data much better. 



The two models give quite different pictures of the 

 climatic effects associated with good spawning success 

 in this Pacific sardine stock (Fig. 5). Correlations with 

 rainfall-related signals were mostly negative for both 

 models. Sea level and sea-surface temperatures were 

 estimated to have mostly negative effects in the cli- 

 mate model, but many of these were estimated as posi- 

 tive in the combined model. Correlations with upwelling 

 variables were all positive in the climate model, but 

 were mixed in the combined model. 



Most correlations with contaminant variables were 

 negative in both models. In the climate model, correla- 



R Rainfall 



O Sea Level 



O Sea Surface Temperature 



U Upwelling 

 • Contaminants 



u ; u 

 u j u 



uj y y u it u u 

 ! °o 



CW 



o 



...» Q .. 



O GD 



^ 6 o 



R Kq 



o 

 o 



RO 



% 



-1 00 -0 75 -0 50 -0 25 00 0.25 0.50 0.75 



Correlation ot Combined' Effect witn Variables 



Figure 5 



Comparison of two regression models of logarithm of spawn- 

 ing success (recruits/spawner) of Pacific sardine Sardinops 

 sagax stock off southern California. Models include effects of 

 stock size and environment. Coordinates of a point are the 

 correlations of the models' explanatory effects (see text) with 

 an explanatory variable. For legibility, only category of vari- 

 able (point I is indicated. Vertical axis: correlations with model 

 estimated on climate data. Horizontal axis: correlations with 

 model estimated on combined climate and contaminant data. 

 Points in first and third quadrants of plane indicate agree- 

 ment between models as to a variable's effect; points in other 

 quadrants indicate disagreement. Stronger negative correla- 

 tions of "combined" model suggest that contaminants have 

 had a negative effect. 



tions ranged from -0.25 to -0.50. In the model with 

 combined effects, the correlations ranged from — 0.25 

 to — 0.95. These strong and consistently negative cor- 

 relations are consistent with the hypothesis that con- 

 taminants affected the dynamics of the Pacific sardine 

 stock, perhaps contributing to its collapse. 



Chub mackerel 



The models of chub mackerel spawning success in- 

 cluded statistically-significant, compensatory stock-size 

 parameters of similar magnitude (Table 5). Compen- 

 sation has been demonstrated in previous analyses of 

 this stock (Parrish & MacCall 1978, MacCall et al. 

 1985), and is apparent in Fig. 3b. 



Including contaminant variables added little to the 

 model. The climate model had lower MSE than the 

 combined model, and its nominal significance level was 

 more extreme (Table 5). One is left with the impres- 



