260 



Fishery Bulletin 88(2). 1990 



correlations, but then have attempted to account for 

 increased intraseries correlation [cf. Drinkwater and 

 Myers (1987), who use a modified form of the Bayley 

 and Hammersley (1946) approach]. 



The goals of the study reported here are to explore 

 potential environmental influences on chinook salmon 

 populations in California's Central Valley, and to dem- 

 onstrate the effective use of some recently developed 

 methods of correlation analysis that are more realistic 

 and conservative than those used previously. We iden- 

 tify potential oceanographic environmental influences 

 on these populations and show that the freshwater in- 

 fluences, which were established by direct survival 

 estimates (USFWS 1987, Kjelson and Brandes 1989), 

 are not detectable from correlation analysis of the 

 available environmental and population data. 



Data and methods 



Environmental data were obtained from four sources. 

 Average monthly streamllow data were obtained from 

 published USGS records for gauging stations at Verona 

 on the upper Sacramento River, Nicholas on the 

 Feather River, and Fair Oaks on the American River. 

 Flow and diversion data for the Sacramento-San Joa- 

 quin delta were obtained from the California Depart- 

 ment of Water Resources DAYFLOW hydrological 

 model. Data on sea surface temperature at the Farallon 

 Islands, southern oscillation index, and tidal height at 

 San Francisco (37.48 N 122.22 W) were obtained from 

 D. Cayan (Scripps Inst. Oceanography, La JoUa, CA 

 92037), and calculated coastal upwelling index at lat. 

 39°N (Bakun 1975) was obtained from A. Bakun 

 (Pacific Fish. Environ. Group, Natl. Mar. Fish. Serv., 

 NOAA, Monterey, CA 93940). 



Population data were taken from the recreational 

 fishery, the commercial fishery, and spawner esti- 

 mates. We used spawner data from the upper Sacra- 

 mento River, the Feather River, the Yuba River, and 

 the American River. In the available data, adults had 

 been separated from jacks on the basis of length. Fish 

 less than 60.7 cm fork length were taken to be jacks 

 and larger fish were counted as adults. For the period 

 1970-86, spawning stock estimates were obtained from 

 Pacific Fishery Management Council (PFMC) reports. 

 For the years 1962-69, total fall-run spawners were 

 taken from Fr\' and Petrovich (1970), and adult spawn- 

 ers were estimated by multiplying the total estimate 

 of fall-run spawners in each stream by the fraction of 

 carcasses classified as adults in the spawning stock 

 surveys. In addition to abundance indices from indivi- 

 dual streams, total adult spawners for the entire Cen- 

 tral Valley were estimated by multiplying the fraction 

 of adults reported in all spawning stock surveys each 



year by the number of spawners in all nms for all rivers 

 combined. 



Commercial and sport catch south of Point Arena 

 were obtained from PFMC reports for the years 

 1971-86, and from L.B. Boydstun (Calif. Dep. Fish 

 Game, Region II, Rancho Cordova, CA 95670, unpubl. 

 data) for the years 1962-70. Commercial effort, 

 measured in thousands of landings, and sport effort, 

 measured in thousands of angler days, were obtained 

 or calculated from the same sources. Fishing effort was 

 used to calculate catch-per-unit-effort (CPUE) in an at- 

 tempt to remove some of the effects of variable fishing 

 effort from the catch data. 



Recruitment of year-classes at the beginning of the 

 commercial fishing season in their second year of life 

 was estimated by deconvolution of the abundance in- 

 dices using the procedure described in Kope and Bots- 

 ford (1988). This involved calculating the contribution 

 of recruitment in each year to the abundance indices 

 in subsequent years using an age-structured catch and 

 escapement model with population parameters esti- 

 mated by separable virtual population analysis of 

 marked hatchery fish (Kope 1987). In addition to the 

 recruitment estimates obtained by deconvolution of the 

 individual abundance series, we estimated recruitment 

 by deconvolving the combined abundance series. This 

 combined deconvolution was obtained by adding to- 

 gether commercial catch, sport catch, and total spawn- 

 ing escapement for each year, and deconvolving the 

 combined abundance series (Kope and Botsford 1988). 



Examination of the different recruitment estimates 

 revealed that the deconvolved spawner series contained 

 a great deal of high-frequency noise which resulted 

 from the marginal stability of the deconvolutions of 

 spawner indices. Because of differences in the relative 

 contribution of each age-class, the deconvolution of 

 total spawners was inherently much less stable than 

 the deconvolution of adult spawners (Kope and Bots- 

 ford 1988). Because recruitment estimates from spawn- 

 ers (adults and jacks) and adult spawners were highly 

 correlated with one another (hence probably represent 

 the same signal, and the deconvolved adult spawners 

 contain less introduced error), only deconvolved adult 

 spawners were used as recruitment estimates derived 

 from spawning escapement. 



No attempt was made to separate natural production 

 from hatchery fish in the spawning escapement. Be- 

 cause of the large contribution of hatchery fish to the 

 spawning runs (cf., Reisenbichler 1986, Dettman and 

 Kelley 1986) and the straying of hatchery fish (Hallock 

 and Reisenbichler 1979, Sholes and Hallock 1979), ac- 

 curate estimation of the natural component of the runs 

 was not possible. 



Before computing correlations with environmental 

 data, we tested for an influence of density on the pop- 



