Jacobson et al.: Depth distributions and time-varying selectivity for various bottom fishes 
323 
Our approach to calculating fishery selectivities for 
commercial bottom trawls is based on the assumption 
that nominal fishing effort data in each stratum from log- 
books is a relative measure of commercial fishing mortali- 
ty. Bias in fishing effort data as a measure of relative fish- 
ing mortality (e.g. due to differences among depth strata 
in average fishing power) would affect our estimates 
of fishery selectivities for commercial bottom trawls. It 
would be better to use a standardized measure of fishing 
effort for each stratum adjusted for differences in season, 
gear, vessel size, engine size, skipper skill, target spe- 
cies, bycatch, or other operational characteristics of the 
vessels (Hilborn and Walters, 1992). We hypothesize that 
this is a minor problem in interpretation of our results, 
however, because the nominal fishing effort data we used 
showed clear and substantial shifts towards deep water 
(Tables 1-2). 
Sensitivity of selectivity estimates to vulnerability 
estimates from field studies 
Vulnerability estimates were the most important uncer- 
tainty in our estimates of fishery selectivities for com- 
mercial bottom trawls (but did not affect estimates of 
depth distributions because they were not used to calcu- 
late depth distributions). We conducted extensive sensitiv- 
ity analysis for each species and determined that a 1-cm 
change in the assumed length at 50% vulnerability (L 50 in 
Table 4) shifted the commercial bottom trawl selectivity 
curves in the same direction by about 1 cm. 
The vulnerability parameters used in our analysis were 
for 4V2-inch mesh which is the current legal minimum 
mesh size in bottom trawls along the west coast, but 
smaller mesh was used by some vessels during earlier 
years (Pacific Fishery Management Council, 1998). Chang- 
es over time in commercial mesh size would affect our 
calculations for early years and small fish. The extent of 
the potential problem is unknown because no vulnerabil- 
ity parameters were available for small mesh gear (e.g. 
3-inch mesh) and because there is no information about 
proportions of total fishing effort by trawls with different 
mesh size during early years. The potential problem is not 
important in estimating selectivity of commercial bottom 
trawls during recent years because recent mesh size regu- 
lations make 4V2-inch mesh standard in the fishery. 
Results from Perez-Comas (1996) were valid estimates 
of selectivities (as defined in his study) but likely underes- 
timate vulnerability (as defined in our analysis) and bias 
our estimates for small fish. Precise definition of terms is 
important in this regard. 
According to Perez-Comas ( 1996), his estimates measured 
“the differential retention of certain sizes of fish after they 
come in contact with the gear” (Gulland, 1983). In our words, 
Perez-Comas measured “the relative probability of capture 
given that a fish entered the mouth of the trawl.” These 
definitions are similar but not identical to our definition of 
vulnerabilities as “the relative probability of capture given 
that a fish is in the path of the trawl.” In particular, Perez- 
Comas’s ( 1996 ) definition differs from ours to the extent that 
fish of different sizes have different probabilities of moving 
« Southern 1978 
10 15 20 25 30 35 40 45 50 
Fork length (cm) 
Figure 8 
Commercial bottom trawl (4V2-inch mesh codend) fishery 
selectivity estimates for shortspine thornyhead (sexes com- 
bined) in the northern and southern subareas during 1978 
and 1996. Selectivity curves for other years were interme- 
diate. Estimates for 1978 are biased owing to commercial 
trawls in the fishery with smaller mesh codends. Smooth 
lines show trends and were fitted to estimates by locally 
weighted regression smoothing (LOESS). 
0.0 Iiiiii — i i i 
10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 
Fork length (cm) 
Figure 9 
Commercial bottom trawl (4'/2-inch mesh codend) fishery 
selectivity estimates for sablefish (sexes combined) in the 
northern and southern subareas during 1978 and 1996. 
Selectivity curves for other years were intermediate. Esti- 
mates for 1978 are biased owing to commercial trawls in 
the fishery with smaller mesh codends. Smooth lines show 
trends and were fitted to estimates by locally weighted 
regression smoothing (LOESS). 
