Heery and Cope: Co-occurrence of bycatch and target species in the groundfish trawl fishery 
43 
Table 3 
Results from free regressions for 5 rare rebuilding rockfish species. Tree 
regressions were conducted for each rebuilding species that did not form 
identifiable clusters in initial cluster analyses in which all species were 
included and were used to develop strata boundaries for additional clus- 
ter analyses. A minimum sample size of 100 tows was required within 
each stratum. Strata selected from the tree regression produced the same 
clustering results that were observed when data from the entire coast 
were analyzed: No clusters containing rebuilding species were evident. 
Species 
Stratification 
Bocaccio 
Canary Rockfish 
Cowcod 
Widow Rockfish 
Yelloweye Rockfish 
46°39.62', 39°53.32', and 38°47.20'N latitude 
47°13.70\ 46°10.72\ and 45°13.19'N latitude 
No suitable strata 
46°05.49' and 40°32.46'N latitude 
47°43.28' and 46°13.53'N latitude 
Sole, English Sole, and Lingcod, but catch of these 3 
species did not serve as an indicator for the potential 
of encountering Canary Rockfish (Fig. 3). 
Next we evaluated whether rebuilding species were 
part of recognizable species assemblages at a finer spa- 
tial resolution. This step was undertaken by conducting 
3 sets of cluster analyses that had been stratified 1) by 
using tree regression, 2) by using 1° latitudinal strata, 
and 3) by departure port. Results from the tree regres- 
sion used to divide tows into latitudinal strata are 
shown in Table 3. Strata selected from the tree regres- 
sion produced the same clustering results that were ob- 
served when data from the entire coast were analyzed: 
no clusters containing rebuilding species were evident. 
The large size of latitudinal strata identified by tree 
regression may explain the similarity of these results 
to the coastwide data results. 
To evaluate species assemblages on an even finer 
scale, the data were divided into smaller latitudinal 
strata (1° intervals). At this spatial resolution rebuild- 
ing bycatch species formed clusters in 1 area, between 
35° and 36° north latitude, where Bocaccio was grouped 
with English Sole (Table 4). PA results indicated that 
the Widow Rockfish was also a part of this assemblage, 
but this finding was not confirmed by HCA. This area 
contained a relatively small number of observations 
with a total of 367 observed tows across all years. In 
other areas, there did appear to be a loose relationship 
between Canary Rockfish and Greenstriped Rockfish 
( Sebastes elongatus). However, this result was not con- 
firmed by all clustering methods and, therefore, was 
not recognized as a meaningful result, particularly giv- 
en that Greenstriped Rockfish is not a targeted species. 
Clusters containing rebuilding species were more 
evident when the data were stratified by departure 
port. Table 5 details the results of cluster analysis for 
each port group separately. Rebuilding species formed 
clusters in several southern port groups. However, the 
sample size for these ports was also 
relatively small. For instance, Bocaccio 
grouped with several slope species, in- 
cluding Sablefish, Dover Sole, and Pacif- 
ic Hake, when caught by 6 different ves- 
sels originating in Santa Cruz and Mon- 
terey. All together, these vessels made a 
total of 294 observed tows (out of 27,162 
total tows coastwide) from 2002 through 
2009. Additionally, Cowcod was associat- 
ed with Pacific Sanddab ( Citharichthys 
sordidus) in the catch of observed ves- 
sels from Avila. These vessels, 7 in total, 
made 108 observed tows throughout the 
entire time series. 
Farther to the north, vessels depart- 
ing from Astoria, Westport, and Ilwaco 
(Fig. 1) tended to catch Pacific Ocean 
Perch with 2 nonrebuilding slope rock- 
fish species: Redbanded Rockfish (Se- 
bastes babcocki) and Splitnose Rockfish 
(S. diploproa). This slope rockfish cluster was distinct 
from the slope species assemblage of Sablefish, Dover 
Sole, and others (Table 5). For Bellingham vessels, the 
slope rockfish assemblage also included Darkblotched 
Rockfish. 
Discussion 
In this analysis, we sought to identify whether there 
were consistently observed associations between tar- 
get and nontarget species in commercial catches moni- 
tored by fisheries observers, who are the source of data 
usually available to fishery managers for bycatch, or 
whether the species composition in the commercial 
catch might be better described as random. We placed 
particular emphasis on the bycatch of rare, but impor- 
tant rebuilding species. Such bycatch of rebuilding spe- 
cies can delay recovery of overfished stocks and limit 
target fisheries. When evaluating data from the entire 
geographic range of the fishery, we found 2 distinct spe- 
cies assemblages that consistently grouped more read- 
ily than did randomly simulated fakies: a deepwater 
slope group characterized by Dover Sole and Sablefish 
and a shallower shelf group dominated by English Sole 
and Petrale Sole (Fig. 2). 
Distinct groups of deepwater slope and shelf species 
have been a consistent finding in studies of species as- 
semblages among west coast groundfish (Gabriel and 
Tyler, 1980; Rogers and Pikitch, 1992; Jay, 1996; Lee 
and Sampson, 2000; Tolimieri and Levin, 2006; Zim- 
merman, 2006; Cope and Haltuch, 2012). The study 
of Rogers and Pikitch (1992) is perhaps the most 
similar to our study, because of their use of observer 
data and similar quantitative methods. The shelf as- 
semblage identified by Rogers and Pikitch (1992) in- 
cluded sanddabs, English Sole, Sand Sole ( Psettichthys 
melanostictus ), Starry Flounder ( Platichthys stellatus). 
