Lo et al.: Application of the continuous egg sampler to estimation of the daily egg production of Sarclinops sagax 
567 
Table 6 
Comparison of egg density and distribution of zero and positive egg counts from CUFES samples with those taken in the laboratory 
and at sea (ship), 1997. CVs are in shown in parentheses. 
Egg density (eggs/min) 
High-density stratum 
Low-density stratum 
Overall 
q 
ship 
4.16 (0.41) 
0.47 (0.45) 
0.213 (0.44) 
laboratory 
4.91 (0.4) 
0.57 (0.45) 
0.211 (0.43) 
ratio 
1.19(0.03) 
1.22(0.05) 
1.20 (0.10) 
Distribution of zero and positive egg counts 
Laboratory 
Ship 
zero 
positive 
total 
zero 
524 
29 
553 
positive 
16 
658 1 
674 
total 
540 
687 
1,227 
1 658 = 130 with exact match + 428 without match. 
Discussion 
Presently the most productive use of the CUFES in the 
DEPM is to use the CUFES as an efficient auxiliary infor- 
mation provider and the CalVET, or another full-water- 
column tow, as the primary sampler. Given our present 
level of knowledge of egg distributions and accuracy of 
predicting them, it would be folly to depend only on the 
CUFES for an estimate of P 0 . Predictive egg-distribution 
models (Sundby, 1983; Westgard, 1989) are promising and 
sometime in the future might make exclusive use of the 
CUFES practical in a DEPM. Exclusive use of the CUFES 
as a sampler in the DEPM also would require detailed 
knowledge of the stage-specific vulnerability of eggs to 
the sampler because a bias seemed to exist for sardine. 
In short, full-water-column tows are essential for accu- 
rately estimating egg production today, but the CUFES 
can greatly facilitate the allocation of such tows. 
The CUFES enhancement of the basic DEPM design 
(Lasker, 1985) may have broad application because the 
DEPM is used world-wide for estimating the spawning 
biomass of sardines, anchovies, and other species of fishes 
(Priede and Watson, 1993;Zeldis, 1993; Lo, 1997). Thus, we 
feel it is useful to discuss some of the new features of the 
DEPM survey that we developed for estimating P 0 with 
the CUFES. 
Critical value for allocation sampling 
To use the CUFES in our DEPM survey design requires 
selecting an egg density that triggers full-water-column 
sampling (CalVET sampling). We used a critical value 
of 2 eggs/min, which was equivalent to 3 or 8 eggs/tow, 
depending on the conversion factor. If the 1997 conversion 
factor (eggs/min=0.25 egg/tow) was correct, we could lower 
the critical value to 1 egg/min, equivalent to 4 eggs/tow. 
This would create a larger high-density stratum and more 
CalVET tows would be allocated. This range of critical 
values (3-8 eggs/tow) was similar to the value (5 eggs/tow) 
used in a stratified sampling design for an anchovy survey 
in Biscay Bay in Spain (Petitgas, 1997). As a result, the 
precision of P 0 and z would likely be improved. Increasing 
the area for the high-density sampling also reduces the 
potential for bias from the assumption of a constant egg 
mortality between strata. On the other hand, lowering the 
critical value diminishes the gain from using the CUFES. 
Clearly, an optimal critical value exists for each species 
and survey area. The critical value can be determined 
prior to the survey or during the survey by using order 
statistics (Thompson and Seber, 1996; Quinn et al., 1999). 
The extent that the critical value can be fine tuned to 
deliver an optimum balance between the CUFES and Cal- 
VET samples for a particular region, species, and season is 
unknown. The large difference in catch ratios between our 
1996 (0.73) and 1997 (0.25) surveys certainly does not sup- 
port the idea of fine tuning. These differences may over- 
state the expected variability for sardine because the ar- 
eas were different; the 1996 samples were taken over a 
very limited portion of the survey area , whereas in 1997, 
the sample pairs were taken where high-density spawning 
occurred (Figs. 1A, 2, and 6). Interestingly, our 1997 esti- 
mate (0.25) is similar to that computed by us for sardine 
off South Africa (van der Lingen et al., 1998). 
Variability in catch ratios 
The extent of vertical mixing of eggs is probably the main 
factor affecting the variation of the catch ratio between 
surveys. Because the selection of the optimal critical value 
for CalVET sample allocation depends upon on the rela- 
tionship between catches in the CUFES and those in the 
CalVET, or the catch ratio, it seems useful to consider what 
