714 



Fishery Bulletin 93(4). 1995 



estimated with some error, given by the standard 

 error of the mean, and because variance terms are 

 additive, this measurement error was subtracted 

 from the total interannual variance in year-class 

 strength prior to calculation of the CV, i.e. 



2 2 2 



o cv = o T0T - a e 



CV = V(e CT " - 1) . 



An estimate of measurement error ( o £ ) was obtained 

 as the mean of the squared standard error estimates 

 (s It 2 ) of the annual index /, averaged over the k years 

 that data were available, 



2 £ 



14- 



t=\ 



Likewise, the total variance in the index (a TOT 2 ) was 

 estimated simply as the sample variance of the in- 

 dex (I t ), 



^■2 

 °TOT 



1 * 



1 * 



iy 



t=i 



For small sample sizes (&<50) and large CV's (>2.0), 

 there is a positive bias in this estimator (Finney, 

 1941). We determined the magnitude of the bias by 

 Monte Carlo simulation (Naylor et al., 1966), and we 

 applied a bias correction term to each CV estimated. 



Shore station sea-surface temperature 



Sea-surface temperature (SST) and salinity are re- 

 corded daily at the University of California Bodega 

 Bay Marine Laboratoiy (BB), the Point Reyes Bird 

 Observatory facility on Southeast Farallon Island 

 (FI), and at the California Department of Fish and 

 Game Laboratory at Granite Canyon (GC) (Walker 

 et al., 1993). SST data from all three sites are gener- 

 ally indicative of hydrographic conditions offshore 

 over the continental shelf (Fig. 1). 



Interannual fluctuations in SST within the cen- 

 tral California study region were estimated by using 

 an analysis of variance (AN OVA) model applied to 

 the shore station data, i.e. 



SSTy^ii + aj+Pj+Yt+e, 



ijk i 



where SST t k is the sea-surface temperature recorded 

 at shore station i U'=BB, FI, or GC) on calendar date 



j {/=1,..., 90} in year k {k=1980,..., 1992}, n is the popu- 

 lation mean SST, and s ijk is a normally distributed 

 error term. Only the first 90 days of the calendar 

 year were included in the analysis because blue and 

 yellowtail rockfish are winter-spawning species 

 (Wyllie Echeverria, 1987) and a measure of the aver- 

 age SST prevailing from birth to completion of the 

 late larval stage was desired. Year effects (y k ) in the 

 model were obtained by calculating population mar- 

 ginal means (i.e. least-square means), providing year- 

 specific estimates of SST at average levels of the a ; 

 and P (see Searle et al. [1980] for further discussion). 



Results 



Annual abundance indices of pelagic juvenile blue 

 and yellowtail rockfishes captured by midwater trawl 

 were quite variable (Table 1; Fig. 2). The CVs of these 

 abundance indices were 1.98 and 1.19, respectively, 

 over the 10-year period from 1983 to 1992. Years of 

 high abundance for both species were 1985, 1987, 

 1988, and 1991, whereas years of low abundance were 

 1983, 1986, and 1992. 



A similar pattern was evident in the data collected 

 by direct underwater observations of recently settled 

 rockfish juveniles in Mendocino and Sonoma Coun- 

 ties (Table 1; Fig. 2), although levels of interannual 

 variation in abundance for these species was some- 

 what greater. Specifically, estimated CV's ranged 



92 93 



Figure 2 



Interannual trends in the abundance of blue [Sebastes 

 mystinus) and yellowtail iSebastes flavidus) rockfishes 

 based on trawl surveys of pelagic juveniles and direct un- 

 derwater observations of settled young-of-the-year fish 

 ( 1983-92). The dashed vertical line shows when the trawl 

 survey was changed from one to three sweeps, (see Meth- 

 ods section). 



