508 
Fishery Bulletin 95(3), 1997 
(Venables and Ripley, 1994). Therefore the initial tree 
was pruned to an optimum number of terminal nodes 
as determined by cross-validation. 
Cross-validation as implemented in S-Plus (Ven- 
ables and Ripley, 1994) uses 90% of the data as a 
training set to grow the tree and test it on the re- 
maining 10%. This procedure is repeated 10 times 
with nonoverlapping test sets. Predictions on the test 
set are done for the initial tree as well as for trees 
pruned to smaller sizes. The resulting deviances are 
computed and plotted against tree size. Deviances 
typically are minimized at an intermediate tree size. 
We chose as optimum tree size the largest size be- 
fore a marked increase in deviance occurred. 
Results 
Rock sole was the most abundant flatfish captured 
in our 1992 sampling (67% of flatfish), as in 1991 
(51% of flatfish). In 1992, a total of 4,625 age-0 rock 
sole (17-60 mm TL) were collected across almost all 
locations, with the highest CPUE in the Sitkinak 
Strait region (Fig. 1A). Age-0 rock sole were found 
mainly near the mouths of bays ± 8-10 km, except for 
a single large catch at the head of Uyak Bay. Age-0 
rock sole were somewhat more abundant with in- 
creasing depth between 0 and 30 m, and were col- 
lected in high numbers to 70 m, although they were 
also found deeper than 70 m. Age-0 rock sole were 
collected in large numbers between 7.5°C and 9.5°C 
and were most often found at salinities of 32.5-33.0 
psu (Norcross et al. 4 ). Rock sole were predominantly 
distributed on sand and mixed sand substrates. Al- 
though found at almost all combinations of depth and 
sand, rock sole were somewhat more concentrated 
in shallow, sandy locations (Fig. 2A). Spearman’s 
rank correlation coefficients (Table 1) indicated that 
rock sole abundance was positively correlated with 
percent sand in the substrate and negatively corre- 
lated with depth, distance from mouth of bay, gravel, 
and mud. Rank correlation was highest with percent 
sand in the substrate. 
Flathead sole increased from 12% of the 1991 catch 
to 18% of the 1992 catch. We captured 1,079 age-0 
flathead sole (23-52 mm TL) during 1992. The dis- 
tribution of flathead sole was more restricted than 
that for rock sole. Age-0 flathead sole were found al- 
most everywhere around the island but were found 
in reduced numbers in Southeast Kodiak (Fig. IB). 
A 0 
04 
£ ° 
CL <9 
< 1 > 
Q 
o 
o 
\ % O o 0 (8 ' ® ^,.991^1 
. : , o • . © % © o cm 
•*. o ° o° • ? . x>° ® 
cP-* • • nW'o# 
• o o co 
. o • ’ * ° ° 
. 98 , -o 
*0 
B 
O 
04 
O 
<9 
0 
0 
v ° 
o ° 9 0 0 . 
*' • . O X 0 . 0 O 0 0 °- 
g? -oo 80 a, 
„ 0 0 0 ^ 
~ 00 0 0 0 
o° n 8 § . 
0 0 ^ 0 q 0 0 
O 
0 20 40 60 80 100 
0 20 40 60 80 100 
Percent sand 
Percent mud 
c 
o 
04 
? 
°-' v ' •••'. ' • 
D 
O 
04 
• Q)Qd 0 R 0 0 °o 
.•9^0 • 8 0 o ^ « 9 n |& 
-C o 
Q. <9 
Q) 
o 
O 
O . 
* A • . • 
° ° . " • * 
O 
<9 
0 
0 
r • • • • 
-10 0 10 20 30 40 
0 20 40 60 80 100 
Distance (km) 
Percent sand 
Figure 2 
Presence (circles) and absence (dots) of (A) age-0 rock sole, (B) age-0 flathead sole, (C) age-0 Pacific 
halibut, and (D) age-1 yellowfin sole plotted against the two “best” discriminator variables. The 
depth axis is plotted with a 120-m limit, and data points occurring between 120 and 180 m are 
plotted at 120 m. 
