Gorfine et al.: Two methods for estimating abundance of Haliotis rubra 
447 
lecting. Indeed, the diver who consistently collected 
more abalone from transects during routine stock 
surveys was not significantly different from other 
members of the dive team when only counts of aba- 
lone were required. Similarly, McShane ( 1995) found 
that by reducing the handling time involved with the 
collection of large numbers of abalone from dense 
aggregations he reduced the operator bias in the es- 
timation of patch frequency. Counting also has the 
advantage of considerably reducing the time needed 
to complete each transect. However, collecting be- 
comes necessary when length-frequency data are 
required because in-situ measurements are impracti- 
cal under the surgy conditions commonly encountered. 
Experience in the radial transect method did not 
have a significant influence on the number of aba- 
lone estimated despite the fact that the number of 
transects completed per diver varyied over a wide 
range (6-438 transects). Differences among divers 
are expected when those who sampled at only a few 
locations are compared with divers who sampled over 
a larger range of abalone populations. This expecta- 
tion is supported by the lower standard deviations 
associated with the mean collections per transect of 
the less experienced divers. 
The standard deviations of the divers’ mean rela- 
tive abundance estimates from radial transects were 
relatively large, especially for those divers who had 
completed many such transects. This large range 
reflects the high spatial variability in abalone den- 
sities both within and between sampling sites which 
is evident from the relatively large proportion of vari- 
ance contributed by location and site effects. During 
the trial stock surveys, variation among sites within 
locations was mostly similar to the variation among 
locations (except for juveniles). Although higher pre- 
cision in abundance estimates may be obtained by 
sampling on a small scale, our results support the 
notion that abalone distributions are variable over 
the range of spatial scales commonly used to census 
abalone populations. Under these circumstances 
stratification of sites may provide little increase in 
the power of population surveys to detect interannual 
changes in abundance. The higher proportion of vari- 
ance between sites in relation to locations with sam- 
pling juvenile abalone was not unexpected because 
of the smaller number of replicates and the greater 
degree of difficulty in observing this cryptic size class 
(Nash et al., 1994). The importance of adopting a 
sampling design that minimizes spatial effects is 
highlighted by the low proportion of the total varia- 
tion attributable to the year effect. The mixed de- 
sign we used provided a compromise between the 
advantages offered by fixed designs in reducing 
within-site variability and those offered by random 
designs in improving the precision of estimates of 
site means. The consequent increase in power over a 
random design has allowed us to economize on the 
number of sites sampled annually. 
The use of transects to estimate the abundance of 
abalone stocks has been criticized (McShane, 1994) 
as a time-consuming approach that does not allow 
for the patchy nature of abalone distributions over 
relatively small spatial scales. However, strip tran- 
sects provide an objective and reproducible approach 
to population surveys that appear to be somewhat 
less affected by research diver differences than are 
observations made against time. Our results dem- 
onstrate that variability among most scientific divers 
in the use of both radial transects and timed searches 
generally has a relatively small effect on variation 
in abalone abundance estimates. This finding sug- 
gests that standardization of abundance data from 
underwater censuses of abalone populations may be 
unnecessary. Because adaptive approaches to the 
setting of TACs require a time series of abundance 
indices, it is unlikely that all scientific dive team 
members will be retained for the required period. 
Thus effects of variation between diver and year will 
confound abundance estimates and interannual 
variation in abundance. The general linear model we 
used to analyze abalone abundance data includes the 
effect of variation among divers. For both methods, 
diver effects were smaller than site effects. Previous 
analyses of timed search estimates involved stan- 
dardization to eliminate diver effects, with the as- 
sumption that all differences between within-site 
replicates were due to diver variation (McShane and 
Smith, 1990). With this approach there is a risk of 
underestimating the variances of mean abalone 
abundances because variation between divers is con- 
founded with intrasite variation. Another problem 
with standardization of data to eliminate diver ef- 
fects prior to analysis is the resulting confidence lev- 
els associated with performing multiple regressions 
between different diver pairs. The greater the num- 
ber of divers the lower the level of confidence. This 
finding contrasts with the higher confidence level 
provided by the general linear models procedure, 
which takes account of diver x site interaction, 
thereby requiring only one comparison among divers. 
The radial transect method is a credible alterna- 
tive to timed searches for the conduct of underwater 
census of abalone. The notion that transect surveys 
are inefficient and time consuming ignores the im- 
portance of how the transects are applied. The num- 
ber of radial transects that can be completed in a 
typical day of sampling compares favorably with the 
number of timed searches that can be completed 
within the same period. When timed searches were 
