Rose et a! : Mortality rates for Chionoecetes opilio, C. bairdi, and Paralithodes camtschaticus after trawls on the seafloor 
47 
tality rate from the control net), resulting in Equation 
3: 
m g+ h = m g + ((1 - m g ) * m h ). (3) 
This equation was solved for m g , resulting in Equa- 
tion 4: 
m g = (m g+ h _ m h ) / (1 - m \)- (4) 
If the cumulative effects of gear impact and handling 
caused additional mortalities, this estimator would at- 
tribute those mortalities to gear effects, resulting in 
overestimated gear-caused mortalities. 
To account for variability due to the combination of 
reflex score assessments, RAMP prediction of mortality, 
and corrections for handling mortality, a randomization 
approach was used for hypothesis testing and estima- 
tion of confidence intervals. A model of the experiment 
was implemented with the Resampling Stats add-in for 
Microsoft Excel (Resampling Statistics, Inc., Arlington, 
VA., http://www.resample.com). 2 RAMP estimators were 
regenerated for each trial by making random binary 
draws for each reflex score category (Urn procedure) and 
by using the sample size and mortality probability for 
that score from the experiment. New probabilities, cal- 
culated from that draw, were then used in the mortality 
estimation procedure for that trial. 
In resampling from the reflex assessments, we used 
each catch as our sample unit, choosing not to assume 
that individual crabs within a catch have independent 
mortality probabilities. To test null hypotheses that 2 
groups of catches (e.g., catches from recapture nets at 
different trawl locations) actually came from the same 
population, the groups were combined and random draws 
were made from that combination, without replacement 
(Shuffle procedure), filling 2 new samples corresponding 
in number to the samples from the original experiment. 
A mortality estimate was generated for each trial by 
using the RAMP and assessment draws. For each test, 
5000 trials were generated, and the proportion of those 
trials with differences greater than the observed esti- 
mate indicated the probability that our result occurred 
from a random process in which the mortality rates for 
both groups were equal. 
Comparisons were made between catches from each 
of the 3 gear areas (center footrope, footrope wings and 
extension, and sweeps) and the control catches to deter- 
mine whether those trawl encounters caused significant 
mortality. Subsequent tests were made for differences 
between the 2 footrope areas and between the sweeps 
and the combined footrope areas. 
Confidence intervals were generated by a similar pro- 
cess, except that samples of the assessment catches for 
each group, including control catches, were randomly 
selected with replacement from the actual catches for 
that group. Handling corrections were applied to mortal- 
2 Mention of trade names or commercial companies is for 
identification purposes only and does not imply endorsement 
by the National Marine Fisheries Service, NOAA. 
ity estimates generated for each gear component, on the 
basis of the control estimate from each trial. Confidence 
intervals (95%) were generated by identification of the 
highest 2.5% and the lowest 2.5% of the estimates from 
5000 trials. 
Effects of sex, size, species, and shell condition on 
mortality rates were examined with logistic regression 
after the effects of each gear component were accounted 
for. Mortality was initially regressed against gear com- 
ponents, and the effects of these other factors were then 
tested against the residual variation. Because logistic 
regression requires binomial outcomes, specific RAMP 
probabilities of death could not be directly applied. 
Where direct observations from holding were not avail- 
able, crab mortality outcomes were scored on the basis 
of whether RAMP probabilities for their mortality were 
less or greater than 50%. Significant effects also were 
tested for interactions of each significant factor with 
gear area. 
Results 
The 159 total tows included 17-21 tows for each spe- 
cies at each recapture position. Between 154 and 991 
crabs from each of the 6 combinations of species and 
sex were assessed after their capture behind each gear 
component and in the control position, and a substan- 
tial range of crab sizes were recorded within each com- 
bination (Table 1). 
Augmentation of the Stoner et al. (2008) RAMP re- 
lationships for the 2 Chionoecetes species by holding 
additional crabs in 2008 had only minor effects on mor- 
tality rate estimates (Table 2) other than to reduce un- 
certainty due to larger sample sizes (Hammond, 2009). 
For all 3 species, injuries varied widely in affected 
body part, type of damage, and severity, and were cor- 
related with both reflex score and mortality rate. Of 
the red king crab with at least one missing reflex (re- 
flex scores of 1 to 6), 96% also had observable inju- 
ries, as opposed to only 5% of those crab with no miss- 
ing reflexes (reflex score of 0). Crabs of all 3 species 
never survived removal of their abdomen or carapace, 
although autotomized legs (dropped off after injury) 
rarely caused fatalities. Crabs with either leg damage 
or carapace cracks normally survived, depending on 
extent, severity, and combination with other injuries. 
Of the 485 surviving red king crab released at the 
end of this study, 482 had all reflexes present upon re- 
lease, including all 14 that initially were missing at 
least one reflex. The 3 crab that were missing a re- 
flex upon release were all missing the eye reflex, had 
been held for 9 or 10 days, and had significant injuries, 
including carapace cracks. Although 25% (122) of the 
surviving crab had detectable injuries, their survival 
through the holding period and vigorous state condi- 
tion upon release indicated a low likelihood of signifi- 
cant later mortalities. 
