266 



Fishery Bulletin 88(2), 1990 



abundance and recruitment is of tiie opposite sign to 

 that expected on the basis of these relationships be- 

 tween the environment and productivity. However, it 

 is possible that because juvenile fish feed on different 

 prey than adult fish, shifts in prey distributions in 

 warm-water years may result in higher growth and sur- 

 vival rates for juveniles. Northward shifts in the dis- 

 tribution of larval clupeids have been documented in 

 warm-water years (Radovich 1959, 1961), and larval 

 anchovies were found closer to shore than usual dur- 

 ing the 1983 ENSO (Brodeur et al. 1985). Larval 

 clupeids are a principal prey of juvenile chinook salmon 

 (Peterson et al. 1982, Pearcy et al. 1985). Data taken 

 off the Oregon coast show no significant difference in 

 feeding by first-year chinook salmon between 1983, a 

 warm-water year, and the other years 1980-85 (Bro- 

 deur and Pearcy In press). 



The results obtained here for freshwater influences 

 during the downstream migration in the first year have 

 greater implications for the general methodology of 

 correlation analysis in fisheries than for the dependence 

 of salmon survival on environment. Other direct obser- 

 vations—the fact that differences in survival between 

 marked fish released above the delta and marked fish 

 released below the delta is correlated with flow over 

 10 years (USFWS 1987)— provide convincing evidence 

 for a negative effect of low flows in the delta. Our 

 results merely underscore the potential iV)r falsely iden- 

 tifying a suspected relationship using standard methods 

 for computing correlations. A dependence of abun- 

 dance on freshwater variables was not detected when 

 intraseries correlation was removed, rather than arti- 

 ficially added, and when it was taken into account in 

 computing correlation coefficients. Fortunately, for 

 this case there have been direct measurements of sur- 

 vival, and we need not rely solely on correlation anal- 

 ysis of abundance and environmental data. 



Although a freshwater effect is not detectable in cor- 

 relations with these freshwater variables, it may pro- 

 vide a partial explanation for the weak correlation seen 

 between population variables and oceanographic con- 

 ditions during the first year. During the spring, the 

 first principle component of oceanographic variables 

 is highly correlated with total delta inflow— flow in the 

 Sacramento, Feather, and American Rivers (0.687, 

 0.577, 0.610, and 0.631, all significant at the 0.01 level 

 using the conservative test). Thus the coi'relation be- 

 tween the first principle component of oceanographic 

 variables during the first spring may be due to occur- 

 rences in freshwater, rather than oceanographic 

 influences. 



The various relationships implied in the results of this 

 study provide a view of remote forcing of the North 

 American marine and terrestrial environment that may 

 be of general utility in examining the effects of envi- 



ronmental variability on other species. Events in the 

 equatorial Pacific appear to drive the oceanographic 

 and meteorological environment in the eastern North 

 Pacific and western North America. Autocorrelations 

 in the SOI (Table 4) indicate that the Equatorial events 

 are spring to winter events (i.e., in column 1, seasons 

 within any spring-to-winter period are correlated with 

 each other, but none are correlated with seasons in any 

 other spring-to-winter period). The rest of Table 4 in- 

 dicates that these events influence conditions in our 

 coastal ocean only in the following winter. This "winter 

 only" aspect of this correlation is consistent with re- 

 sults relating an index of atmospheric pressure condi- 

 tions on the west coast of North America, the Pacific 

 North American index to events in the ecjuatorial 

 Pacific (Horel and Wallace 1981). Mysak (1986) at- 

 tributed this "winter only" aspect of the equatorial/ 

 midlatitude teleconnection to the recjuirement for 

 strong westerlies between the equator and midlatitudes 

 in order to advect the atmospheric Rossby waves 

 responsible for the teleconnection to the northeast. 

 These westerlies exist only during the winter. The 

 covariability among our three oceanographic variables 

 (i.e., the fact that UPW is as important as SLH and 

 SST in the loading on the first principle component) 

 suggests an atmospheric teleconnection with equatorial 

 ENSO rather than a wave propagating pf)leward from 

 the Equator (cf. Emery and Hamilton 1985, Mysak 

 1986). These variables are also related to winter rain- 

 fall in central California. For example, annual rainfall 

 in Sacramento is significantly correlated witli the first 

 principle component and the upwelling index in winter. 

 Flows through the Sacramento-San Joa(iuin Delta the 

 following spring are in turn correlated with that rain- 

 fall, as well as winter oceanographic conditions. This 

 remote forcing on a global scale and the associated 

 covariability of marine and terrestrial variables further 

 confound the identification of causal meclianisms from 

 these kinds of data. 



However, even direct measures of survival versus 

 environmental variables do not necessarily identify 

 mechanisms unequivocally. Covariation between flow 

 rates, diversions, and reversed flows in the lower San 

 Joaquin River continue to obscure the true cause of 

 lower loin size in the San Joaquin River system (Kjelson 

 and Brandes 1989, USFWS 1987). 



The differences in results between standard methods 

 and the new methods used here are a rough measure 

 of how much more conservative they are. At the 0.10 

 level, for example, the more conservative method of 

 computing the variance of the estimate of correlation 

 coefficient yielded significant correlations roughly half 

 as often as standard methods (Tables 1-3). These 

 results, together with simulations that show that the 

 combination of equations (1) and (2) yield the prescribed 



