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



315 



preceding values as necessary (Table 2). The extrapo- 

 lated 1981 values, although necessary for the princi- 

 pal-component analyses, had very little influence on 

 later calculations. 



The reconstructions of Summers et al. (1988) com- 

 prise the best available dataset, and the only one on 

 this scale, describing contaminant loadings off south- 

 ern California in recent decades. Because of this, veri- 

 fication is difficult. Prager & MacCall (1990) found 



that the reconstructions agreed fairly well with some 

 recent estimates of metals loadings made by the South- 

 ern California Coastal Water Resources Project 

 (SCCWRP), which has monitored fluxes of metals from 

 southern California sewage outfalls since 1971 (Konrad 

 1989). However, only a small portion of total metals 

 loadings are in sewage; this may have caused the dis- 

 crepancies noted for some metals by Prager & MacCall 

 ( 1990). We also compared reconstructions with a study 

 of metals deposition in anaerobic sediments in the 

 Santa Barbara basin (H. Schmidt & C. Reimers, Scripps 

 Inst. Oceanogr., La Jolla CA, pers. commun.). How- 

 ever, the sedimentary record, much less precise than 

 the reconstructions, was unable to discern any clear 

 trends or patterns in metals since 1932. 



Explanatory data: Climate 



A large suite of climate data (Table 3) for the years 

 1920-84 was compiled by Prager & MacCall ( 1987a,b,c), 

 who described the data sources and processing in de- 

 tail. Most of the raw data were provided by Dr. An- 

 drew Bakun (Pac. Fish. Environ. Group, NMFS South- 

 west Fish. Sci. Cent., Monterey CA). 



We detrended most of the sea level (SL) and sea- 

 surface temperature (SST) series to remove rising 

 trends with time. Values for 1983 were not used in 

 computing trend lines, to prevent this extreme El Nino 

 year near the end of the series from distorting the 

 results. Only the series of SST in San Diego did not 

 exhibit a significant trend, and accordingly that series 

 was not detrended. 



Seasonal effects were removed by standardizing each 

 variable to zero mean and unit standard deviation (SD) 

 by month of year. For example, the January values 

 (1920-84) of Los Angeles rainfall were standardized 

 as a group to mean 0.0 and SD 1.0. Standardization 

 was performed separately for each month and time- 

 series. 



The climate dataset had a few missing values that 

 we replaced with estimates. We used the BMDP proce- 

 dure AM (Dixon et al. 1983) to estimate these values 

 by stepwise regression on the available data. Reason- 

 ableness of the estimates was verified by simulation 

 and by comparison with the nonmissing data (Prager 

 & MacCall 1990). 



Principal-component analyses 



Many of the explanatory variables contained redun- 

 dant information. For example, sea-surface tempera- 

 ture and sea level exhibit very similar patterns over 

 time, as both are related to El Nino conditions and the 

 flow of the California Current. Another example, noted 

 above, is that many contaminant time-series are tightly 



