Lo et al.: Application of the continuous egg sampler to estimation of the daily egg production of Sordinops sagax 
565 
Table 4 
Estimates of daily egg production (P 0 \ number of eggs/0.05 m' 2 /day at age 0), daily instantaneous mortality rate (z) for the high- 
density stratum (66,841 km 2 ) and low-density stratum ( 107,255 km 2 ), P Q for the entire survey area ( 174,096 km 2 ), the catch ratio 
of eggs/min in the low-density stratum to eggs/min in the high-density stratum ( q ), their standard errors (SE), estimated bias, and 
the bias-corrected estimates for the unweighted and weighted nonlinear regression from 1000 iterations of bootstrap simulation. 
High-density stratum 
Low-density stratum 
Entire survey area 
P 0,l 
SE (P 01 
) 2 
SE (z) 
P 
-*0,2 
SE (P Q 2 ) 
P 0 
SE (P 0 ) 
Q 
SE ( q ) 
Unweighted nonlinear regression 
survey 
5.04 
1.28 
-0.21 
0.16 
1.06 
0.52 
2.57 
0.72 
0.211 
0.09 
bootstrap 
mean 
5.10 
1.60 
-0.29 
0.28 
1.10 
0.53 
2.60 
0.85 
0.22 
0.08 
SE 
1.60 
0.72 
0.12 
0.11 
0.48 
0.28 
0.71 
0.34 
0.09 
0.03 
CV 
0.31 
0.46 
-0.40 
0.38 
0.45 
0.52 
0.27 
0.41 
0.41 
0.40 
bias 
0.06 
0.32 
-0.08 
0.12 
0.03 
0.01 
0.03 
0.13 
0.009 
-0.01 
bias corrected 
4.98 
0.96 
-0.13 
0.04 
1.04 
0.51 
2.54 
0.58 
0.202 
0.10 
Weighted nonlinear regression 
survey 
4.76 
0.86 
-0.35 
0.050 
1.004 
0.45 
2.43 
0.51 
0.211 
0.09 
bootstrap 
mean 
3.50 
0.72 
-0.27 
0.056 
0.730 
0.33 
1.80 
0.41 
0.22 
0.08 
SE 
1.60 
0.35 
0.10 
0.025 
0.400 
0.21 
0.76 
0.20 
0.09 
0.03 
CV 
0.46 
0.49 
-0.36 
0.440 
0.550 
0.63 
0.42 
0.48 
0.42 
0.40 
bias 
-1.26 
-0.14 
0.08 
0.006 
-0.274 
-0.12 
-0.63 
-0.10 
0.009 
0.01 
bias corrected 
6.02 
1.00 
-0.43 
0.044 
1.278 
0.57 
3.06 
0.61 
0.202 
0.10 
linear regression was unwarranted in our case. One pos- 
sible reason is that the standard errors (SE) of egg produc- 
tion in each age group were between 1.3 and 2.8, whereas 
the SE for yolksac larvae was 0.16 (Table 3). Although 
the weighted regression should be used when the varianc- 
es of data are unequal, we believe that there is too large 
a disparity between variance of eggs and yolksac larvae, 
and too much weight was assigned to yolksac larvae for a 
weighted nonlinear regression. 
The spawning biomass was computed from the daily 
egg production from the 1997 survey, and the historical 
daily specific fecundity (number of eggs/gram of biomass) 
was 23.55 (Macewicz et al., 1996) assuming the daily spe- 
cific fecundity was the same as that for 1994-96. Sardine 
spawning biomass in 1997 would be 379,940 t 3 for an area 
of 174,096 km 2 (50,868 nmi 2 ) from San Diego to San Fran- 
cisco. No variance of B s was computed because no variance 
of the number of eggs per population weight (g)/day was 
available. 
Comparison of results from the 1997 CUFES and 
DEPM survey with results from a conventional 
DEPM survey 
We compared the results of the 1997 CUFES and DEPM 
design with those of a conventional DEPM ( 1994), in which 
only CalVET samples were taken, to illustrate how the 
3 The spawning biomass would be 359,280 metric tons if P 0 of 
2.43/0.05 m 2 from weighted nonlinear regression was used. 
CUFES allocation design may affect the performance of a 
DEPM survey of Pacific sardine. We believe the compari- 
son is instructive, even though the two surveys differed 
somewhat in area and population size; the conventional 
1994 survey covered a larger area than the 1997 survey 
(380,175 km 2 vs. 174,096 km 2 ), and, the total biomass 
of sardine was smaller in 1994 than in 1997 (111,493 t 
(Lo et. al., 1996) vs. 379,940 t, respectively). In both sur- 
veys staged eggs and yolksac larvae from CalVET samples 
were used in the calculation of P 0 . 
An obvious difference in the results of the two surveys 
was that only 11% (74/684) of CalVET samples were pos- 
itive for sardine eggs in the 1994 conventional survey, 
whereas in the 1997 survey 72% (102/141) were positive 
(Table 5). These results indicate that CUFES was effec- 
tive in allocating CalVET samples and thereby reducing 
ship-time costs per survey mile. The coefficients of varia- 
tion (CV) for the estimates of P 0 were similar: 0.22 for the 
conventional survey compared with 0.27 for the CUFES- 
based DEPM. Thus, variance penalty for using the ratio 
estimator q did not greatly diminish the benefit in using 
the CUFES in the DEPM. This simple statistical compari- 
son, however, does not reveal the greatest potential bene- 
fits in using a CUFES. The allocation of CalVETs would be 
most useful when the population is at a lower level, as it 
was in 1994, because at such levels one must cover a large 
survey area to assure an unbiased estimate, but the popu- 
lation is probably concentrated in a very small fraction of 
the area where CalVET samples will be allocated. In addi- 
tion, the high-resolution maps of the spatial distribution 
