Laman et al.: Correlating environmental and biogenic factors with abundance and distribution of Sebastes alutus in Alaska 
285 
et al., 2013). Juvenile Pacific ocean perch have been 
observed in close association with sponges and corals 
on the bottom (Carlson and Straty, 1981; Rooper et 
al., 2007) and appear to favor more spatially heteroge- 
neous habitats (Du Preez and Tunnicliffe, 2011). Adult 
Pacific ocean perch are known to form schools on or 
near the seafloor (Krieger, 1993; Brodeur, 2001; Han- 
selman et al., 2012) where we typically collect them 
over areas of softer seafloor substrata, such as sand or 
gravel, during our surveys. Adults have been observed 
in association with sea whip forests (Brodeur, 2001), 
although they appear, on the basis of our model results, 
to be less benthically oriented than juveniles. The ap- 
parent affiliation of Pacific ocean perch with biogenic 
structures over trawlable habitats in the Aleutian Is- 
lands could potentially modify their distribution and 
abundance patterns on a local scale within the context 
of the broadly acting biological and oceanographic pro- 
cesses that determine larger scale patterns. 
All our models retained biogenic structures with 
morphological features capable of increasing struc- 
tural heterogeneity in the otherwise low-relief, traw- 
lable habitats, but none retained the Sp predictor, 
the presence-absence factor compositely representing 
all sponges. Most of the sponges and corals retained 
in our models are erect forms (e.g., V and F sponges 
and primnoid corals), but even the nonerect forms 
(e.g., G and Gp sponges) are known to be epizoic on 
larger, erect sponge forms (Stone et al., 2011). In pre- 
vious studies that showed associations of Sebastes spe- 
cies (Heifetz, 2002; Krieger and Wing, 2002; Du Preez 
and Tunnicliffe, 2011) and Pacific ocean perch (Rooper 
and Boldt, 2005; Rooper et al., 2007) with sponges and 
corals, the epibenthic invertebrates were considered 
composite categories. The results of our study refine 
our understanding of the relationship between Pacific 
ocean perch distribution and abundance and the struc- 
tural heterogeneity provided by biogenic structures 
from these larger composite groupings. Our results in- 
dicate that habitat heterogeneity, and vertical relief in 
particular, can affect Pacific ocean perch distribution 
and abundance. 
Although studies have indicated that the presence 
of structure-forming epibenthic invertebrates can lead 
to enhanced abundance and biodiversity of associated 
animals (e.g., Du Preez and Tunnicliffe, 2011; Beazley 
et al., 2013; Knudby et al., 2013), Tissot et al. (2006) 
concluded that associations of fishes with sponges and 
corals do not necessarily imply functional relationships 
between these groups of organisms. The co-occurrences 
between Pacific ocean perch and biogenic structures 
that we modeled in the Aleutian Islands may derive 
from facultative relationships (e.g., structural refugia 
and prey availability) or could simply result from a 
convergence of conditions that favor the presence of 
both the fish and epibenthic invertebrates. The under- 
lying mechanisms leading to the patterns observed in 
our study are an important area for future study. 
Generalized additive models relate response vari- 
ables to dependent model variables through additive 
and unrestrictive smooth functions, making them well 
suited for modeling typical interactions of species with 
environment (Hastie and Tibshirani, 1986; Maravelias, 
2001; Guisan et al., 2002). The GAMs also provide a 
data-defined assessment of the shape of the response 
of a species to independent variables (Maravelias and 
Papaconstantinou, 2003). The anticipated nonlinear 
relationships of Pacific ocean perch distribution and 
abundance with their environment made GAMs well 
suited to provide a tool that was more informative than 
traditional regression techniques. 
One drawback to the use of GAMs is that they can 
over-fit the data, resulting in unrealistic models with 
limited predictive power (Kim and Gu, 2004; Wood, 
2006). To minimize the risk of over-fitting, we con- 
strained the df available to smoothed continuous pre- 
dictor variables in the model and successfully validated 
our GAM predictions of juvenile and adult presence 
with a data set external to the modeling effort (i.e., 
the data from the 2012 Aleutian Islands trawl survey). 
We conclude that the GAMs parameterized in this 
study were fairly accurate predictors of juvenile occur- 
rence but were better at predicting adult occurrence. 
Furthermore, our validation results indicate that these 
models were robust predictors of the distribution and 
abundance of Pacific ocean perch juveniles and adults 
in the Aleutian Islands in subsequent years. 
Our approach of modeling presence and absence and 
conditional abundance independently is a technique 
intended to cope with zero-inflated and over-dispersed 
data and that is common to abundance surveys (Mc- 
Cullagh and Nelder, 1989; Barry and Welsh, 2002). Pro- 
cesses that influence distribution and those that affect 
abundance do not have to be the same. By modeling 
the 2 response variables independently, we were able 
to determine that the models for Pacific ocean perch 
distribution and conditional abundance shared more in 
common than not. For example, the best-fitting models 
for both life stages across both model classes all share 
the D,T and Long, terms in common, but for the pres- 
ence-absence GAMs adults and juveniles also share 
the SI and Vt terms. By comparison, there is no over- 
lap in the suite of biogenic structures retained in the 
juvenile presence-absence and conditional abundance 
GAMs, a result that may indicate that the local con- 
ditions leading to enhanced probability of encounter- 
ing juvenile Pacific ocean perch in our trawl tows are 
not the same as those leading to increased abundance 
when this species is present. 
There is ample evidence that rockfishes occur com- 
monly in areas that are untrawlable with our present 
net (Carlson and Haight, 1976; Zimmermann, 2003; 
Rooper et al., 2007; Rooper et al., 2011). Underwater 
camera and submersible observations indicate that 
some untrawlable areas are havens for biogenic struc- 
tures and fishes (Rooper et al., 2007; Stone et al., 2011). 
In some instances, catch rates for Pacific ocean perch 
varied with tidal velocity (i.e., juvenile and adult oc- 
currence). Changing catch rates could result from tidal 
current regimes that affect distribution and abundance 
