Gorfine et al.: Two methods for estimating abundance of Haliotis rubra 
441 
Table 1 
Analysis of variance in density estimates, made from radial transect collections, among fifteen sites nested within three locations 
and among four divers for each of three size classes of H. rubra (type IV SS). ns = nonsignificant; * = P<0.10; ** = P<0.05; *** = 
PcO.Ol). co 2 = relative magnitude of variance estimate (%). MS = mean square. 
Class 
Source 
df 
MS 
F 
P 
c o 2 
Significance 
Juvenile 
Location 
2 
0.14 
2.42 
0.1312 
14 
ns 
Diver 
3 
0.05 
1.92 
0.1593 
14 
ns 
Location x diver 
4 
0.04 
1.36 
0.2822 
3 
ns 
Site (location) 
12 
0.06 
1.69 
0.0904 
41 
* 
Diver x site (location) 
20 
0.03 
0.75 
0.7631 
25 
ns 
Residual 
63 
0.04 
Prerecruit 
Location 
2 
2.36 
19.24 
0.0002 
44 
*** 
Diver 
3 
0.06 
1.10 
0.3723 
0 
ns 
Location x diver 
4 
0.08 
1.32 
0.2936 
2 
ns 
Site (location) 
12 
0.12 
3.56 
0.0003 
37 
*** 
Diver x site (location) 
21 
0.06 
1.70 
0.0475 
18 
** 
Residual 
81 
0.03 
Postrecruit 
Location 
2 
2.60 
9.98 
0.0028 
43 
*** 
Diver 
3 
0.33 
3.16 
0.0461 
16 
** 
Location x diver 
4 
0.14 
1.30 
0.3027 
3 
ns 
Site (location) 
12 
0.26 
1.95 
0.0393 
27 
* 
Diver x site (location) 
21 
0.10 
0.78 
0.7390 
11 
ns 
Residual 
90 
0.13 
ing previously used the other method, performed both 
techniques to census abalone populations at each 
study site. Abalone counts in four radial transects 
and four timed collections of 10-min duration were 
made by each diver at each site so that the effects of 
divers and methods were fixed and the design bal- 
anced. Timed searches were performed after transect 
counts in each instance to avoid disturbance due to 
removal of abalone during timed searches. 
Separate analyses for each method were made with 
general linear models to determine the significance 
of the effects of diver and site on abundance esti- 
mates. A combined analysis was not possible because 
of the different base sampling units for each method. 
Where Cochran’s test indicated significant hetero- 
geneity of variance, data were log-transformed prior 
to analysis to achieve homoscedasticity (Winer, 1971). 
Diver x site interaction was omitted from the analy- 
ses presented in Table 1 because preliminary analy- 
ses showed this effect to be nonsignificant. 
Trial stock surveys 
The effects of divers on the detection of differences 
in abundance were further investigated during trial 
surveys of substocks at Gabo Island and Sandpatch 
Point near Mallacoota and Sticks Reef near Altona 
in Port Phillip Bay (Fig. 1). These surveys were aimed 
at determining how well variation among several 
locations was detected with this method when it was 
applied over a spatial scale comparable to that used 
in routine stock surveys. 
Five sampling sites were randomly selected at each 
location, and for each site, three out of five divers 
were selected to undertake the sampling depending 
on their availability and fitness to dive. Two to three 
sites at each location were sampled during each day 
of the surveys. At each site, each diver made three 
replicate radial transect collections of emergent 
prerecruit and postrecruited abalone. Each diver also 
collected juvenile abalone (<80 mm) in one of their 
three assigned transects by searching cryptic habi- 
tat and turning over small boulders. All abalone col- 
lected were brought on board the research vessel, 
and the maximum shell length of each was measured 
to the nearest mm. 
Analysis of variance of the mixed linear model was 
used to test the effects of locations, sites nested within 
locations, and divers on abundance estimates. The 
SAS® general linear model (GLM) procedure was 
used (SAS Institute, Inc., 1989). Separate analyses 
were performed for three size classes of abalone (ju- 
venile <80 mm, 80-mm < prerecruit < LML, and 
postrecruit > LML). Data were converted to number 
per m 2 . 
The relative treatment magnitude, omega squared 
(to 2 ), for each effect in the GLM was calculated 
(Keppel, 1991). 
