76 
Fishery Bulletin 99(1 ) 
Tabie 2 
Estimates of the variance components a 2 d (between divers) and a\ 
(within divers) of the natural logarithm of relative accuracy (esti- 
mated length/true length), for experienced scientific divers, novice 
divers, and the stereo-video system (90% confidence limits shown in 
parentheses). The value of o 2 d is taken to be zero for the stereo-video 
system. 
Between divers Within divers 
Experienced 
divers 0.015 (0.006,0.067) 0.022 (0.017,0.030) 
Novice divers 0.016 (0.008,0.063) 0.023 (0.017,0.031) 
Stereo-video 
system 0.00020 (0.0001,0.0005) 
of 15% with 90% power for blue cod when only 
one fish is being recorded per sample, 30 samples 
per site need to be recorded with the stereo-video 
system, 70 with experienced scientific divers, and 
71 with novice scientific divers. When 30 fish are 
recorded per sample only 10, 27, and 28 samples 
are required, respectively (Fig. 4). 
Influence of a 
With stereo-video measurements, an increase in 
the value of a has little effect on the number of 
samples required to achieve a high level of power. 
When a is set at 5%, 10%, and 20%, we required 
10, 9, and 7 samples per site, respectively, to detect 
a 15% change in the mean length of our popula- 
tion of blue cod with 90% power (Fig. 5). To achieve 
a similar level of power with experienced scien- 
tific divers 27, 22, and 17 samples needed to be 
recorded per site, whereas 28, 23, and 18 samples 
were required for novice scientific divers. 
Discussion 
Our research suggests that the power to detect 
changes in the mean length of three common spe- 
cies of fish from coastal waters of New Zealand 
will be higher when length estimates are made 
by a stereo-video system rather than by experi- 
enced or novice scientific divers. In addition, the 
level of power to detect these changes will differ 
between species, depending on the natural vari- 
ability in mean length between different sites and 
locations. Managers and environmental research- 
ers need to be aware of this issue when selecting 
suitable species of fish to be included in monitor- 
ing programs. For some species of fish (e.g. red 
cod), the variability in mean length between sites 
may be so great that even with a stereo-video 
system, unrealistically large numbers of samples 
need to be taken per site to detect changes in the 
order of 30% with high statistical power. 
A stereo-video system may have many advan- 
tages in monitoring programs, particularly where 
few fish are recorded per sample. Fewer samples 
need to be taken per site with a stereo-video sys- 
tem to obtain an equivalent level of power com- 
pared with the number of samples needed with 
experienced scientific divers (Fig. 2), thus both 
time and money are saved in the field. 
Even though calibration procedures are recom- 
mended and used by some researchers (GBRM- 
PA, 1979; Bell et al., 1985; Polunin and Roberts, 
1993; Darwall and Dulvy, 1996), interdiver vari- 
ability and diver measurement error may still in- 
validate comparisons of the changes in the mean 
length between sites and over time. Consequently, 
it is important that the level of precision and ac- 
curacy of length estimates is stated in reports and 
