Rooper: An ecological analysis of rockfish assemblages in the North Pacific Ocean 
9 
0.4 0.5 0.6 
Overlap 
Figure 5 
Frequency of co-occurrence of species in trawl hauls versus combined 
multinomial overlap index. Data are shown for the top 10 species- 
subgroups in terms of total catch per unit of effort for the Gulf of 
Alaska and Aleutian Island trawl surveys, and each dot represents a 
comparison of species-groups 0i = 90). 
different from more commonly used meth- 
ods where trawl survey catches or stations 
with similar components are clustered 
together. More commonly used approaches 
define assemblages by comparing patterns 
of catch by species in trawl hauls (i.e., 
Weinberg, 1994; Williams and Ralston, 
2002; Zimmermann, 2006). Although 
these methods are highly effective for de- 
termining patterns in catches, they typi- 
cally are based on complex analyses such 
as nonmetric multidimensional scaling, 
principle components, and cluster analy- 
ses that can make interpretation difficult. 
Determining the scaling method to apply 
to catch data for abundant versus rare 
species, the patchy distribution of some 
species, and inherent differences in catch- 
ability (such as between small and large 
fish of the same species) are all problems 
that must be addressed with these meth- 
ods. By first defining the relationship of 
a species to environmental variables, and 
then comparing the parameters of that 
relationship to parameters for other spe- 
cies or life history stages, this analysis method may 
avoid some of those potential pitfalls. For example, the 
absolute abundance or catchability should not matter 
in the identification of the correct assemblage member- 
ship for a species. If the proper distribution of a species 
along a depth gradient is known from the trawl data; 
the species will be placed within a group of species 
with similar depth distributions regardless of its total 
abundance. 
The issue of whether trawl survey data can be used to 
determine the underlying relationship of rockfish spe- 
cies to large-scale environmental gradients may have 
some limitations. Less sampling effort was directed at 
shallow inshore depths (0-50 m) than at greater depths; 
therefore shallow-water species were likely under-repre- 
sented in the catches. Although this analysis included 
three environmental gradients, there are obviously more 
variables needed to fully describe the niche dimensions 
for any of the rockfish species. One important feature 
that was omitted from this analysis was the effect of 
small-scale habitat features on rockfish distribution. 
Variability in rockfish species assemblages on a small 
scale is often related to the local habitat, where higher 
habitat heterogeneity is correlated with higher diversity 
and abundance (Matthews, 1990; Stein et al., 1992; 
Yoklavich et al., 2000). The data used in this analysis 
was collected only on trawlable ground, and large tracts 
of untrawlable area were unsampled. Species composi- 
tion and abundance can be starkly different between 
trawlable and untrawlable locations (Matthews and 
Richards, 1991). Even within trawlable areas, habitat 
characteristics such as presence of epibenthic inverte- 
brates can be correlated with increased catches of some 
life history stages of rockfish, such as juveniles that 
seek out complex habitat features (Rooper and Boldt, 
2005; Rooper et al., 2007). Larger-scale phenomena, 
such as patterns in prey productivity and the effects of 
local currents on rockfish distribution, were also absent 
from this analysis and can certainly affect the distribu- 
tion of those species that prey on planktonic organisms. 
Geographical position was used as a proxy for these 
large-scale phenomena and other unknown variables 
influencing rockfish distributions. By pooling the data, 
I did not take into account interannual changes in 
geographical position due to climate events, such as El 
Nino or the Pacific decadal oscillation shifts. The ef- 
fects of El Nino events are typically exhibited through 
changes in water temperature and there was only a 
weak response of the species to temperature gradients, 
and major climate shifts were not detected in Alaska 
during the years examined. Knowledge of the effects of 
climate shifts and local habitat features would further 
improve future assemblage analyses. 
Because these analyses were based on rockfish re- 
lationships with environmental variables, they should 
result in predictable species assemblages useful for 
ecosystem-based management. For example, species 
that co-occur in trawl catches due to overlapping dis- 
tributions with environmental variables would be likely 
to experience similar fishing mortalities. Additionally, 
based on these assemblage analyses, marine protected 
areas could be designed for specific depth and geo- 
graphical areas that would protect portions of rockfish 
populations. A series of marine protected areas has 
been suggested for shortraker and rougheye rockfishes 
in the Gulf of Alaska as a first step towards spatial 
management (Soh et al., 2000). The spatial and depth 
separation of juveniles and adults in many of the spe- 
cies examined here could also provide information to 
implement spatially based management systems for 
