FISHERY BULLETIN: VOL. 70. NO. 3 



proportion of positive stations was 87% "near- 

 shore" and 68% in the "nearshore-excluded" re- 

 gion with respect to anchovy larvae. 



In Table 7 I have compared the annual esti- 

 mates of anchovy larvae using the quarterly re- 

 gional census estimate of the entire region with 

 the same estimate using the sum of the near- 

 shore and nearshore-excluded segments of the 

 region. The latter estimate is 8 % lower than the 

 regional census estimate, but it is not likely to 

 be a bias since of the 10 annual estimates, the 

 single estimate exceeds the partitioned estimate 

 5 times while the reverse is true an equal number 

 of times. The Spearman rank coefficient of cor- 

 relation is 0.85. Also, a comparison of the an- 

 nual estimates from 1951 to 1959 regional cen- 

 sus estimates and those from a census estimate 

 by Ahlstrom (1967, Table 2) shows a Spearman 

 rank correlation coefficient of 1.00. Thus, we 

 may conclude that errors of the kind involving 

 nearshore gradients in the incidence and inten- 

 sity of anchovy spawning and larval survival, 

 while important for some applications, do not 

 measurably affect the regional census estimates. 



One may reasonably ask whether estimates 

 generated from monthly cruises are comparable 

 to larval abundance estimates from a single 

 cruise in each quarter. In Table 7 I have com- 

 pared the annual estimates of anchovy larvae 

 using all regular occupied stations within the 

 quarter and a similar estimate using only Jan- 

 uary, April, July, and October, with February 

 used in 1957 because the January cruise was in- 

 complete. The Spearman rank coefficient for 

 the comparison is 0.90: the mean value of the 

 estimates from monthly cruises is 10% below 

 the estimates from one cruise per quarter but 

 in 5 years the monthly derived estimates exceed 

 the quarterly and in 5 years the reverse is true. 

 Thus, I conclude that differences which may 

 arise from comparing estimates from quarterly 

 and monthly cruises are not large enough to af- 

 fect this study. 



Technical errors should be relatively small and 

 affect the catches of sardine and anchovy simi- 

 larly. For example, the factor a,, the area of 

 mouth of the net is usually known to within 5%, 

 the factor bi is modified by the flow through the 

 mouth of the net so that the length of tow is 



50r 



20- 



10 



5- 



£2- 



8 9 10 



LENGTH (mm) 



13 



Figure 6. — The size frequency of sardine larvae cap- 

 tured in 1951-60 in the southern California inshore (clear 

 dot, solid line), offshore (black dot, small dash), and 

 seaward (triangle, large dash) regions. The size fre- 

 quency effect is attributed to offshore transport (see 

 text). 



underestimated by 13-15%: an error as great 

 as 5-10% results from the flowmeter being cen- 

 tered in the backwash of the bridle apex hard- 

 ware (Smith and Clutter, 1965; Tranter and 

 Smith, 1968; Mahnken and Jossi, 1967). 



Two biases, which have yet to be evaluated, 

 are likely to be important. In cold water, the 

 larvae may tend to grow more slowly. Thus the 

 regional census estimate for a cold quarter or 

 cold year could prolong the period for which 

 the larvae are vulnerable to sampling. Simi- 

 larly, larvae may persist without food for ex- 

 tended periods yielding the same kind of error 

 mentioned for temperature. No estimate has 

 been made for either bias for anchovy or sardine 

 larvae. 



All the sardine and anchovy larvae collected by 

 the CalCOFI net have been subjected to trans- 



860 



