142 
Fishery Bulletin 95(1), 1997 
CL 
<D 
cr 
c n 
q3 
cl 
CL 
TO 
C 
c n 
o 
o 
z 
4 
2 
0 
30 
25 
20 
15 
10 
5 
0 
N. Kaneohe S. Kaneohe Kailua 
North Kaneohe 
Station numbers of spatial replicates 
Figure 4 
Mean (filled circle) and range (hollow circles) of east Oahu snapper station 
reliability test. The top graph is the 1994 multicanyon analysis and the 
bottom graph is the multiyear assessment ( 1992-94). The “T” on each graph 
represents the location of each canyon’s axis or “trough.” Numbers of rep- 
licates (Rep.) at each station are represented with bar graphs at the top of 
each graph. Maps on the right of each graph indicate the areas covered by 
the data. 
dance was significantly correlated with only the clay- 
silt (<0.0625 mm) fraction (Spearman’s r s =0.35, 
P<0.001) (Table 2). The greatest abundance of clay- 
silt occurred in an area just northwest of the north 
Kaneohe canyon trough; at Kailua the abundance 
was less than half that at Kaneohe, and high con- 
centrations spread southeast of the canyon trough. 
Proximity of point sources of drainage was associ- 
ated with snapper abundance, i.e. video index of 
snapper abundance and distance to discharge were 
significantly negatively correlated (r s =-0. 18, P<0.05). 
The weakness of the relationship resulted from the 
failure to consider the effect of bottom relief in the 
comparison. This result indicated the need for a 
model that considered the variables together. 
Modeling of snapper aggregations. 
The stepwise backward regression evaluated the rela- 
tive importance of the 3 habitat variables found sig- 
nificant in the univariate analysis: 1) escarpment- 
type relief; 2) clay-silt (<0.0625 mm) sediment frac- 
tion; and 3) proximity of coastal discharge. The in- 
teraction of discharge with the presence of clay-silt 
Table 2 
Spearman rank order correlation coefficients and probabil- 
ity values for snapper abundance with sediment particle 
size. In all comparisons sample size = 211. 
Sediment size 
fraction (mm) 
Correlation 
coefficient r s 
Probability 
value P 
>2.000 
-0.0426 
0.54 
0.350-2.000 
-0.0932 
0.178 
0.149-0.350 
-0.0922 
0.184 
0.063-0.149 
0.0809 
0.244 
<0.063 (clay-silt) 
0.3555 
<0.001 
was also assessed. All variables except clay-silt were 
retained by the model (P<0.01; Table 3). Reasons for 
the model’s exclusion of clay-silt will be discussed 
later. The model correctly predicted overall presence 
(>5) or absence (<5) of snapper aggregations for 68% 
of the video drops. The model predictions of pres- 
ence (79% [>5]) were roughly balanced by those for 
absence (60% [<5]) (Table 4). Ranked snapper abun- 
dance was interpolated by using all video drops to 
