Watters and Dick: Length distributions of Sebastes spp. off central California 
299 
Table 4 
Values for 10%, 50%, and 90% total length quantiles (cm) and results of a chi-square two-sample 
test in comparing length distributions of 4 species of rockfish (Sebastes spp.) from submersible 
surveys in low-relief (<25 cm), trawlable habitat and from trawl surveys conducted off central 
California (36-37°N latitude) during 2003-2009. Length data were estimated to the nearest 5 cm 
during submersible surveys and to the nearest 1 cm during trawl surveys. For the chi-square test, 
length data from trawl surveys were binned to 5-cm increments (bin as midpoint) to match the 
format of data from submersible surveys. 
Species 
10%, 50%, 90% 
length quantiles, 
untrawlable habitat 
10%, 50%, 90% 
length quantiles, 
trawlable habitat 
Chi-square 
two-sample test 
Greenspotted rockfish 
10, 15, 30 
14, 26, 36 
138.9, df=9, P<0.001 
Greenstriped rockfish 
5, 15, 25 
17, 23, 29 
574.01, df=7, P<0.001 
Vermilion rockfish 
35, 45, 50 
42, 47, 52 
16.58, df=7, P<0.02032 
Canary rockfish 
15, 30, 45 
37, 42, 50 
73.905, df=10, P<0.001 
to cause the reduced numbers of small sizes we ob¬ 
served in the trawl survey data. 
An important consideration for our study was the 
accuracy of visually estimated fish lengths from the 
submersible survey. Trawl length data are in-hand 
measurements to the nearest cm, whereas submersible 
length data are visually estimated underwater to the 
nearest 5 cm with the aid of paired lasers. Yoklavich 
et al. (2007) conducted a study to address the error as¬ 
sociated with visual estimates of fish length from the 
Delta submersible and found lengths were underesti¬ 
mated by 1.1 cm on average. Given that the trawl sur¬ 
vey data were binned to 5-cm increments for plotting 
and the chi-square test, and maximum lengths from 
the submersible survey were similar or greater than 
those from the trawl survey, we did not consider the 
relatively small amount of error associated with visu¬ 
ally estimated length data to have contributed greatly 
to the differences found in our study. 
Before our comparisons of lengths, we surmised that 
lengths from the submersible survey would be more 
similar to lengths from the trawl survey for green'’pet¬ 
ted and greenstriped rockfishes than for canary and 
vermilion rockfishes, on the basis of known habitat 
associations of these species that would make them 
more or less available to the trawl survey. To some 
extent, this assumption held true. Greenspotted rock- 
fish, which associate with a wide variety of high- and 
low-relief habitats (Yoklavich et al., 2000; Love et al., 
2002; Laidig et al., 2009), had similar length distribu¬ 
tions and mean lengths between the surveys when all 
fish, regardless of habitat, were compared; the distri¬ 
butions were significantly different, however, because 
of the large amount of length data from the submers¬ 
ible survey. Canary and vermilion rockfishes associate 
strongly with high-relief rock habitats (Yoklavich et 
al., 2000; Love et al., 2002; Laidig et al., 2009), had 
the most dissimilar length distributions between the 
surveys for all fish regardless of habitat, and had the 
least amount of length data from the trawl survey. Our 
result for greenstriped rockfish was somewhat surpris¬ 
ing because, given that this species associates most 
commonly with low-relief trawlable habitats (Yoklav¬ 
ich et al., 2000; Love et al., 2002; Jagielo et al., 2003; 
Laidig et al., 2009) and had the greatest amount of 
length data (601 measurements) from the trawl survey, 
we did not expect to find such strong dissimilarity in 
the length distributions between the surveys. However, 
the trawl survey selectivity curve from the stock as¬ 
sessment for greenstriped rockfish (Hicks et al., 2009) 
correctly assumes that smaller fish were not sampled 
by the trawl survey. 
Given that adults of many rockfish species are known 
to associate with high-relief rocky habitats, for assess¬ 
ments based on trawl survey data, it is often assumed 
that selectivity for the survey is “dome-shaped,” i.e., 
availability of larger fish to the survey may decline be¬ 
yond a given size (Dick et al., 2011; Taylor and Wetzel, 
2011; Hamel et al., 2013; He et al., 2015). One mecha¬ 
nism for this pattern could be ontogenetic movement 
into untrawlable habitat (Love and Yoklavich, 2008). 
Although limited in spatial and temporal extent com¬ 
pared with stock assessments based on trawl survey 
data from the entire U.S. west coast and multiple years, 
our results for these 4 species suggest that the major 
difference between size compositions from the submers¬ 
ible and trawl surveys may be a reduced frequency of 
smaller individuals in the trawl survey. The estimated 
selectivity curves in the assessments of greenspotted 
and greenstriped rockfishes appear to account for this 
