Zeldis et a I.: An estimate of biomass of Hoplostethus atlanticus 
591 
mate of egg production but could cause its precision 
to be overestimated. No attempt was made to cor- 
rect for this bias because the data were judged to be 
inadequate to estimate an autocorrelation structure 
(which would need to have both spatial and tempo- 
ral terms). Picquelle and Megrey (1993) drew the 
same conclusion for pollock egg surveys. For the 
present study, a minimum spacing of 1,000 m was 
imposed when allocating the stations to the strata. 
This spacing should have gone some way toward 
minimizing the effects of spatial autocorrelation. In 
an explicit study of autocorrelation in anchovy egg 
catches, Smith and Hewitt (1985) found that spatial 
autocorrelation diminished rapidly at spatial scales 
of 2,000 m and that temporal autocorrelation dimin- 
ished at 0.5 h for 1-day-old eggs. Given that the mini- 
mum distances and times between samples for the 
egg stages used in this study were of this order and 
that the egg catch rates were highly variable, over- 
estimation of precision was likely to be minor. 
Use of the DFRM required the assumption that 
the 11 trawls used to estimate D representatively 
sampled spawning females on Ritchie and North 
Hills. Although it is clear that spawning female bio- 
mass is highly aggregated in the Ritchie Hill area 
(Fig. 2), it is not known to what extent females are 
randomly distributed within this small area, with 
respect to ovarian stage. Therefore, it was useful to 
look at a much larger set of fecundity data collected 
on these hills in 1995 by N.Z. Ministry of Fisheries 
(MOF) scientific observers working on commercial 
vessels (Zeldis, unpubl. data). Samples were taken 
from 47 trawls made with four vessels (range of 7— 
16 trawls per vessel) throughout the entire spawn- 
ing season. These tows were long, typically travers- 
ing much of the ridge between North Hill and the 
south side of Ritchie Hill (Fig. 2) and covering the 
850-900 m depth range of orange roughy spawning. 
The fecundity reduction rate from these 1995 data 
(uncorrected for turnover) was 965 eggs/(kg x day) 
(CV=0. 11 ), which was not significantly different from 
the uncorrected estimate from 1993 in the present 
study, 787 eggs/(kg x day) (CV=0.11; Table 4). This 
rate showed that widespread and intensive trawling 
in this area would yield an estimate of D similar to 
that from the less intensive research trawling of the 
present study, indicating that the less intensive 
trawling accurately represented the spawning dy- 
namics of the population. 
The reliance of the DFRM on within-spawning sea- 
son trawl data makes it susceptible to bias due to 
turnover, which causes an underestimate of fecun- 
dity reduction rate. In the present study, it appeared 
that spent females left the survey area during the 
period when fecundity reduction was measured, such 
that only 0.11 of remaining females were spent 21 
days after the start of spawning. A similar pattern of 
low proportion spent was seen in the 1995 Ritchie 
Bank scientific observer data described above, with 
proportion spent < 0.20 about 23 days after the on- 
set of spawning. In contrast, in recent NIWA orange 
roughy egg production surveys done at East Cape in 
1995 10 and at the “Graveyard” area on northern 
Chatham Rise in 1996 11 (Fig. 1), spent proportions 
were between 0.60 and 0.70 about 20 days after the 
onset of spawning. This suggested that turnover was 
less prevalent at these latter sites, an effect which 
may be related to the fact that although commercial 
fishing was intensive on Ritchie Bank during the 
1993 and 1995 spawning seasons, there was no com- 
mercial fishing during the 1995 and 1996 spawning 
seasons at East Cape and the “Graveyard.” Signifi- 
cantly, the value of D at East Cape was 1,036 eggs/ 
(kg x day), with no correction for turnover, which was 
similar to the turnover-corrected value for the 
present study at Ritchie Bank of 1,106 eggs(kg x day) 
( an estimate is not yet available for the “Graveyard”). 
For this reason, the turnover-corrected fecundity 
reduction rate estimated for Ritchie Bank in 1993 is 
considered to be more reliable than the uncorrected 
estimate and also a reasonably reliable estimate of 
the true rate of decline in Ritchie Bank population 
seasonal fecundity. 
Comparision with other studies 
The first use of the DFRM was by Lo et al. (1992; 
1993) to estimate the biomass of a deepwater 
pleuronectid flatfish, Dover sole ( Microstomus 
pacificus), which spawns between 600 and 1,500 m 
depth on the continental slope of western North 
America. The model used in the present study to 
describe the daily fecundity reduction was simpler 
than that of Lo et al. (1993). It assumed that the 
fecundity per fish weight was a linear function of 
time only (fish weight was considered as an addi- 
tional predictor but did not significantly improve the 
fit). Lo et al. (1993) assumed that both total fecun- 
dity of active females and the fraction of active fe- 
males (their E f and G ( ) were linear functions of time 
and fish weight. The former model was used for two 
reasons. First, in calculating fecundity reduction, 
there seemed no need to treat active and inactive 
females separately. Second, in the absence of any 
evidence of lack of fit, the rule of Occam’s razor sug- 
gested using the simpler model. 
11 Grimes, P. J. 1996. Voyage Report, TAN9608 (Part II). NIWA 
unpublished voyage report held in NIWA Library, Greta Point, 
Wellington, New Zealand, 4 p. 
