Parrish et al.: Nursery habitat in relation to production of Pristipomoides filamentosus 
145 
sence of snapper aggregations in the 
logistic model. This finding supports 
the hypothesis that structural relief 
or its associated community repre- 
sents conditions less favorable or 
more hazardous to the snappers 
(greater interspecific competition, 
risk of predation, etc.) (Johannes, 
1978; F. A. Parrish, 1989). Expanses 
of uniform sediment bottom are obvi- 
ously an important substrate feature. 
The relative scarcity of this habitat 
observed at depths <60 m at least 
partly explains the absence of snappers 
on the shallower (30-60 m) grounds. 
Proximity to point sources of drain- 
age and its relationship with the 
presence of clay-silt sediment can 
explain much of the snappers’ long- 
shore distribution. Work with first- 
year juveniles of species of Pagrus 
has demonstrated that substrate and 
associated water flow are important 
to habitat selection (Francis, 1995). 
Improved availability of food has 
been proposed as a reason for fish 
demonstrating habitat preferences 
(Sudo et al., 1983). Distributions of 
sediment particle sizes such as clay- 
silt have been shown to enhance the 
localized distribution of certain 
benthic invertebrate infauna (Fegley, 
1988). A favorable localized sediment 
composition might contribute to an 
enhanced forage base for juvenile fish 
(Tito de Morais and Bodiou, 1984). 
However, these longshore variations 
in clay-silt abundance are simply in- 
dicative of the longshore differences 
in coastal water flow that disperse 
the flocculent clay-silt. The highest 
fraction of clay-silt is found at the 
seaward end of the north Kaneohe 
channel, where snapper abundance 
is high and bay drainage is most con- 
centrated. The density of fish in 
Kailua is greatest near the wastewa- 
ter outfall, where the clay-silt frac- 
tion is lowest. The outfall introduces 
and increases the frequency of drift- 
ing materials to the area, similar to 
the flow of natural drainage sources, 
but without creating a clay-silt dis- 
persion field. For this reason, the lo- 
gistic model excluded the variable 
