512 
Fishery Bulletin 95(3), 1 997 
C 
aJ 
> 
a> 
O 
Rock sole Flathead sole 
19.0 8 1 5.2 2.4 7.20 1.70 0.95 
Pacific halibut 
870 4.10 094 
Yellowfin sole 
89 5.1 2.0 
Tree size (number of terminal nodes) 
Figure 4 
Cross-validation plots (deviance versus tree size) used to prune regression trees of catch per unit of effort on 
five habitat parameters. See text for further explanation. 
sediment and depth selected by linear discriminant 
analysis as an important factor in determining the 
distribution of juvenile flathead sole. 
Distance from the mouth of the bay and depth were 
the best predictors of halibut CPUE (Fig. 7) The de- 
viance of the pruned tree was 0.587 compared with 
0.384 for the initial tree. Highest CPUE values oc- 
curred at stations less than 40 m deep and more than 
2.9 km outside the mouth of bays (10 fish/10-min 
tow). Very low abundances or no halibut were found 
at stations more than 7.9 km up the bay (0.13 fish/ 
10-min tow). Intermediate CPUE values were found 
at 61 stations near the mouth of bays (-2.9 km to 7.9 
km) which had high bottom temperature (>9.0°C) on 
sand or mixed sand substrates (2.9 fish/10-min tow). 
This confirmed our earlier finding that halibut tend 
to remain outside or near the mouth of bays in water 
less than 40 m deep on sandy substrates. 
The most important variables used in predicting 
yellowfin sole abundance were depth, sediment, and 
distance. Deviance was increased from 0.442 for the 
initial tree to 0.895 for the pruned tree. The first split 
separated 100 stations less than 28 m deep from 69 
stations deeper than 28 m (Fig. 8). The deeper sta- 
tions had a very low mean CPUE (0.17 fish/10-min 
tow), and yellowfin sole were absent at 64 of the 69 
stations that were deeper than 28 m. The shallow 
stations had a low mean CPUE (0.82 fish/10-min tow) 
if the substrate type was pure gravel, sand, or mud, 
or had mixed gravel sediment, whereas stations on 
mixed mud sediments had medium to high abun- 
dances of yellowfin sole (9.0 fish/10-min tow). The 
53 shallow stations on mixed mud substrates were 
further split by distance, indicating that the highest 
CPUE values occurred near the heads of long bays. 
Thus yellowfin sole tended to be concentrated in very 
shallow locations on mixed mud sediments near the 
head of bays. This finding agreed with results of the 
linear discriminant function in its identification of 
depth and sediment as important factors, but fur- 
ther added distance from the bay mouth as a third 
important factor. 
