Zeldis et al.: An estimate of biomass of Hoplostethus atlanticus 
579 
tions optimally in a survey of size N stations, n / sta- 
tions were allocated to each stratum layer such that 
rij c< A ; M ; 
and 
where A • = the area of each stratum layer. 
Allocation was done in proportion to the strata means 
(Mj) because they were highly correlated with the 
strata standard deviations and were probably esti- 
mated more reliably than the standard deviations 
(Francis, 1984). To estimate values of N that would 
yield a desired coefficient of variation for egg abun- 
dance estimates, the standardized counts in each 
stratum layer were randomly sampled with replace- 
ment (bootstrapped) to estimate M-, where the num- 
ber of samples taken from each stratum was n. .. The 
survey mean egg abundance combined across all 
strata, E, was estimated as 
This procedure was repeated 500 times and the mean 
of the 500 survey estimates was taken as the egg 
abundance estimate. The standard deviation of the 
500 estimates divided by the mean was taken as the 
coefficient of variation of the egg abundance estimate. 
This analysis suggested that the optimal alloca- 
tion would have stations allocated to the central, 
middle, and outer strata in ratios of 1.5:0.38:0.25, 
respectively. It also suggested that 400 stations would 
provide adequate precision (CV=0.15) in the egg 
abundance estimate. Therefore, for the AEPM design, 
five 80-station subsurveys were planned with sta- 
tions allocated to strata in the above ratios. 
Using simulations, we found that by occupying sta- 
tions within each stratum in an order which mini- 
mized steaming distance between stations, before 
moving to a new, randomly chosen stratum, about 
50% less steaming time would be involved than by 
occupying stations in completely random sequence in 
each subsurvey. This procedure would be done, how- 
ever, at the cost of variable (and possibly long) periods 
of no coverage of each stratum between subsurveys and 
could be a serious drawback if spawning intensity var- 
ies significantly and rapidly (over a few days) during 
