590 
Fishery Bulletin 95(3), 1997 
Sources of error 
In egg-production surveys, the trajectory of the plank- 
ton net during deployment and retrieval is usually 
assumed to be straight (and often, vertical). The 
analyses presented here (Appendix 1) suggest that, 
for the present survey, 1) there was significant hori- 
30000 
25000 
20000 
t 
55 15000 - 
ao 
PJ 
10000 
5000 - 
05 Jun 10 Jun 15 Jun 20 Jun 25 Jun 30 Jun 05 Jul 10 Jul 15 Jul 
Date 
Figure 8 
Weight-specific fecundity (eggs/kg spawning female: R i ) from trawls 
prior and during the DFRM sampling period (circles and squares, 
respectively, Table 2). Solid line is a weighted linear regression 
fitted for DFRM trawls. Dotted line is a regression fitted when it 
was assumed that proportion active was overestimated by 39% on 
4 July (see text). 
zontal movement of the net during deployment ow- 
ing to drag from the warp (which was caused by ship 
drift), and 2) the plankton net followed a curved tra- 
jectory during hauling. The curved trajectory meant 
that less water was filtered in deeper than in shal- 
lower ocean layers of equal thickness and that larger 
correction factors were required for younger (deeper) 
eggs in order to standardize the egg counts to 
eggs/m 2 . The effect of using curved, rather 
than straight, trajectories was to increase the 
estimated production rate by 36% (from 8.0 to 
10.9 billion eggs/day). Recent egg-production 
survey work, using a digitally recording flow- 
meter system developed at NIWA (Grimes 9 ), 
has shown that 1) there often are spurious 
revolutions of the flowmeter on the downcast, 
negating the use of conventional flowmeters 
and that 2) more water is filtered at shallower 
depths than at deeper depths, justifying the 
assumption of a curved net trajectory. There- 
fore, it is likely that by allowing depth varia- 
tion in the estimates of the amount of water 
filtered, a major improvement was made over 
the assumption of a straight net trajectory. 
The precision of the planktonic egg produc- 
tion estimate was influenced by damage to 
early stage eggs (Zeldis et al., 1995; Grimes et 
al. 3 ). In the present study, the inability to stage 
most eggs less than 21.8 h old caused a greater 
reliance on older egg stages to estimate pro- 
duction. However, the number of older egg 
stages that could be used for the esti- 
mation was limited because eggs older 
than stage 10 (28.4 h) were subjected 
to advection toward and through the 
southwestern boundary of the survey 
area as they aged. Thus, the original 
strategy of making the survey area 
large enough to retain all of the eggs 
up to 36 h old was defeated, and rela- 
tively few stages were available for a 
mortality estimate. Clearly, the com- 
bination of nearly concurrent informa- 
tion on ocean circulation (from CTD 
and drogued buoy) and on the drift pat- 
terns of the eggs themselves provided 
useful corroborative information for 
quantitative decisions about the egg- 
age range available for mortality esti- 
mation, when eggs were subject to ad- 
vection. 
Precision estimates for the egg- 
abundance data may have been biased 
by potential autocorrelation among the 
data, which would not bias the esti- 
