126 



Fishery Bulletin 105(1) 



RS, 



2 



I 



a=l 



S*-^a.2.(+2+-Ra.3./+3) 



(S,.2 + S,.3)/2 



(3) 



where t = the year juvenile sockeye salmon were sam- 

 pled (f=2000, 2001, 2002); 



a = the freshwater age (1.0 or 2.0); 



R = is the total number of returning adult sock- 

 eye salmon to Bristol Bay after year t; and 



S = the total number of spawners in Bristol 

 Bay that contributed to the juvenile salmon 

 population during year t. 



For instance, freshwater juvenile sockeye salmon (ages 

 1.0 and 2.0) sampled during 2000 came from cohorts 

 spawning during 1998 (age 1.0) and 1999 (age 2.0) and 

 returned to Bristol Bay during 2002 as adult salmon at 

 age 1.2 and 2.2 in 2002 and at age 1.3 and 2.3 in 2003. 

 The numbers of returning adult and spawning Bristol 

 Bay sockeye salmon were estimated from brood-year 

 return information provided by ADF&G. 



Annual indices of juvenile sockeye salmon abundance 

 (lA) were defined as 



IA,=[SA/ina)C,, 



(4) 



where SA = the estimated survey area (189,000 km^); 



rna = the mean area sampled by a trawl haul 



during the survey (distance traveled during 



the tow multiplied by the width of the net); 



and 



the mean number of juvenile sockeye 

 salmon caught during year t ('?=2000, 2001, 

 2002). 



C, 



This formula would give the abundance of juvenile sock- 

 eye salmon in the survey area if we assumed that catch- 

 ability with our midwater trawl was 1, (i.e., all fish in 

 front of the net were caught). Because this is unlikely, 

 we treat our estimates as an index rather than as actual 

 abundance. In fact, our juvenile sockeye salmon abun- 

 dance indices were less than the resultant adult returns 

 in some years, indicating that catchability of the net 

 was much less than 1. One study (Shuntov et al., 1993) 

 where larger surface trawl gear was used to sample 

 juvenile salmon indicated that catchability of juvenile 

 salmon was 0.3. We therefore divided our abundance 

 indices by 0.3, although we still considered these values 

 to be indices. 



An index of juvenile sockeye salmon marine-stage 

 survival rate (IMS) was estimated by 



sampled (^=2000, 2001, 2002), and a is freshwater age 

 (age 1.0 or 2.0). 



These survival rate indices were correlated with the 

 mean length of juvenile fish collected during the cor- 

 responding first year at sea. Because the mean date 

 for juvenile sockeye salmon sampled for length differed 

 among years, we adjusted fish lengths to provide a 

 standardized length using September 1 as the standard 

 date. Adjusted mean fish lengths were calculated by as- 

 suming three different daily growth rates: 1) mm/day, 

 representing no daily growth at sea; 2) 0.3 mm/day, the 

 lower end of published growth-rate ranges for juvenile 

 Pacific salmon; and 3) 1.7 mm/day, representing the 

 upper end of the ranges (see Fisher and Pearcy, 1988, 

 1990; Fukuwaka and Kaeriyama, 1994; Orsi et al., 

 2000 for daily growth-rate ranges for juvenile Pacific 

 salmon). 



Results 



Analyses of adult scale data 



Examination of the autocorrelation and partial auto- 

 correlation functions for Kvichak River freshwater age 

 groups 1 and 2 and the Egegik River freshwater age- 

 group-1 MSWl univariate time series indicated that 

 these time series had a constant mean and variance. 

 For the Egegik River freshwater age-group-2 MSWl 

 growth index, the sample autocorrelation and partial 

 autocorrelation functions indicated that a lag-1 autore- 

 gressive parameter was appropriate and the estimate 

 of the parameter was significant (^test, P<0.01). Coef- 

 ficients of variation were less than 4% for the MSWl 

 growth-rate indices for each freshwater age group, and 

 thus confirmed the univariate model results that these 

 time series varied little over time. By comparison, the 

 coefficients of variation for the time series of returns per 

 spawner for each freshwater age group were between 

 70% and 135%. 



The MSWl growth index was not significantly related 

 to survival in any of the LTF models except for Egegik 

 freshwater age group 1 (Table 2). Parsimonious univari- 

 ate models were reasonable explanations of survival for 

 both river systems and age groups, having values of 

 SBC nearly as low as the "best" models. The sample 

 autocorrelation and partial autocorrelation functions 

 indicated that a lag-1 autoregressive parameter was 

 appropriate for the all of the univariate survival rate 

 time series models. The estimates of the lag-1 autore- 

 gressive parameter were positive for all of the univari- 

 ate models. 



IMS, = s^ X 100, 



lA, 



(5) 



where R is defined above in Equation 3, /A, is defined in 

 Equation 4, t is the year juvenile sockeye salmon were 



Analyses of data from juveniles collected by trawling 



The distribution of juvenile sockeye salmon along the 

 eastern Bering Sea varied among years (Fig. 3). During 

 2000 and 2001, 75% of the total catch of juvenile sock- 

 eye salmon occurred south of 56°N, within the middle 

 domain and south within the stratified waters near the 



