Prager and MacCall Contaminant and climate effects on spawning of three pelagic fishes 



321 



sion that adding the contaminant variables merely 

 muddied the picture. 



The two models give similar pictures of the forces 

 that appear to be associated with variability in spawn- 

 ing success of chub mackerel (Fig. 6). We consider the 

 climate variables first. Both models showed moderate- 

 to-strong positive correlations of spawning success with 

 sea-surface temperature early in the year, moderate- 

 to-positive correlations with sea level early in the year, 

 and weak-to-moderate correlations with upwelling later 

 in the year. Correlations with rainfall-related variables 

 were weak. 



The models were also similar in correlations with 

 contaminant variables. Correlations with metal load- 

 ings were weak-to-moderate, but of different signs for 

 different metals; this suggests chance (noncausal) cor- 

 relations. Much of the signal in the metal loadings 



-0.8 -0.4 4 8 



Correlation of "Combined" Effect with Variables 



Figure 6 



Comparison of two regression models of logarithm of spawn- 

 ing success (recruits/spawner) of chub mackerel Scomber 

 japonicus stock off southern California. Models include ef- 

 fects of stock size and environment. Coordinates of a point 

 are the correlations of the models' explanatory effects (see 

 textl with an explanatory variable. For legibility, only cat- 

 egory of variable (point) is indicated. Vertical axis: correla- 

 tions 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 agreement between models as to a variable's effect; 

 points in other quadrants indicate disagreement. Strong agree- 

 ment between models suggests that addition of contaminant 

 data adds little to this model. 



occurred in the late 1960s and beyond, when the chub 

 mackerel stock was declining from overfishing and then 

 recovering (MacCall et al. 1985); this coincidence could 

 have induced noncausal statistical correlations. Corre- 

 lations with organochlorines were negative. In inter- 

 preting this, it is relevant that spawning success in 

 chub mackerel is highly autocorrelated and appears 

 periodic (Fig. 3b). The organochlorine abundances are 

 also very highly autocorrelated, usually consisting of a 

 single peak. This could easily lead to correlations like 

 those observed, even if the observed contaminant lev- 

 els did not affect spawning success. 



Although several contaminants correlate very highly 

 with spawning success, the correlation patterns are 

 nearly identical for the two models, one of which con- 

 tains no information on contaminants. We interpret 

 this to indicate that any contaminant effects are 

 indistinguishable from effects of climate variables 

 that fluctuate similarly over time. The strong agree- 

 ment between the two models also suggests that 

 recruitment-based contaminant analyses of this stock 

 may be exceptionally prone to Type-II error, i.e., fail- 

 ure to detect any contaminant effects that might actu- 

 ally be present. 



Parrish & MacCall (1978) also explored the effects 

 of climate on chub mackerel spawning success, but 

 comparison of this work and theirs is difficult for two 

 reasons. Parrish & MacCall (1978) examined many 

 explanatory variables not included in this study In 

 addition, changes in the estimated maturity schedule 

 and the VPA-based stock and recruitment information 

 (Prager & MacCall 1988) have resulted in a markedly 

 different, as well as considerably extended, stock and 

 recruitment series. Nonetheless, many of our estimated 

 correlations agree with those of Parrish & MacCall 

 (e.g., the positive effects of sea-surface temperature in 

 the south, and the upwelling index at 30°N). Whereas 

 Parrish & MacCall found that spawning success had a 

 negative correlation to sea level at La Jolla, we esti- 

 mated positive correlations with sea level at La Jolla 

 and Los Angeles from January to June and indepen- 

 dence in the later months. 



Statistical evaluation of contaminant effects 



In a preceding section we developed three statistical 

 criteria for admission of possible contaminant effects. 

 The sign test required 12 negative correlations between 

 16 "deleterious" contaminant variables and the explana- 

 tory effect of a model using the combined explanatory 

 data. Correlations for models of all three stocks were 

 mostly negative. Models of northern anchovy (Fig. 5) 

 and chub mackerel (Fig. 6) each contained 11 negative 

 correlations, but neither result was sufficient to reject 



