82 
Fishery Bulletin 120(1) 
fisheries, gears, vessel types, and years are represented 
in the database at higher rates because of differential 
rates of observer deployment (full coverage versus par- 
tial coverage or interannual changes in coverage; see 
AFSC and ARO’). Combining data from different sam- 
pling strata without correcting for coverage rates will 
result in biases toward those areas of the fishery with 
higher observer coverage. 
Because of these differential coverage rates and the 
limited information on fishing effort from the unobserved 
portion of the fisheries, it is not possible to use the num- 
ber of sampled hauls (nor the number of sampled trips) 
and the total number of hauls (nor the total number of 
trips) to obtain a direct estimate of the total number of 
interactions that occur for all fisheries. A proxy for the 
ratio of sampled hauls to total hauls has been used to esti- 
mate mortality of marine mammals by assuming that the 
weight of groundfish from sampled hauls (as determined 
by the observer during sampling), as a fraction of the total 
groundfish weight landed and reported on fish tickets for 
the stratum (year, fishery, area, and marine mammal 
species), is equal to the ratio of the number of sampled 
hauls to the total number of hauls in the stratum. This 
approach is typical for estimating incidental mortality in 
an observed fishery (Breiwick, 2013); however, we have 
not used this approach in this analysis for several reasons 
detailed later in the text. 
Our focus was to examine only interactions that 
occurred on observed hauls, not to estimate the total 
number of the various interaction types in each fishery. 
Interactions recorded by observers may occur on hauls 
that are not scheduled for sampling but during which the 
crew notifies the observer of an interaction. The number 
of hauls monitored for interactions includes these hauls 
in addition to hauls randomly selected to be sampled 
(and monitored for interactions with mammals), and as 
a result the number of monitored hauls may be greater 
than the number of sampled hauls, increasing the nom- 
inal monitoring rate relative to the rate associated with 
the randomized sampling design. Although the number 
of animals involved is recorded by the observer, some 
animals present might not be observable (e.g., swim- 
ming below the surface or otherwise out of sight of the 
observer). In addition, multiple animals may be involved 
in any interaction, and some animals may be included 
in counts associated with different interaction types. 
Lastly, interactions are not necessarily independent of 
each other on a haul or within a trip. For example, in 
cases where deterrence is used to decrease depredation 
by mammals on the catch, feeding on catch or feeding 
on discards may also be recorded (i.e., for a single haul, 
3 interactions, each of a different type and each involving 
multiple animals, may be recorded). 
For the reasons given in a previous paragraph, we have 
not used ratio estimates based on groundfish weights to 
make inferences about interactions of marine mammals 
with fishing gear for all fishing events in Alaska. We have, 
however, summarized the frequency of interactions, tak- 
ing into account the changes that occurred in observer 
coverage throughout the years by comparing the percent- 
age of monitored hauls in which interactions were docu- 
mented for the full-coverage and partial-coverage strata 
separately, although sampling rates in partial-coverage 
strata have varied between years after 2013 and varied by 
gear types in 2016 (AFSC and ARO’). 
The percentage of hauls in which interactions with 
killer whales occurred was computed as the number 
of hauls in which at least one interaction (of any type) 
was recorded divided by the number of hauls monitored 
by observers (multiplied by 100), an estimate of hauls 
with killer whale interactions per 100 monitored hauls. 
This estimate is inclusive of hauls that were not sam- 
pled but during which the observer or the vessel crew 
noted an interaction. These percentages were computed 
for each year and for the partial- and full-coverage strata 
separately. By presenting the percentage of hauls with 
interactions for each sampling stratum, the analysis also 
accounts for differences in sampling intensity. Interac- 
tions with killer whales that were recorded for trips that 
were not associated with a specific haul were not included 
in this analysis (39 interactions of the 3245 interactions 
recorded since 2008). 
About the tables and figures of this article in which 
numbers of interactions are used, it should be noted that 
the number of interactions with killer whales is more than 
the number of hauls with such interactions. As noted here 
previously, this difference results from some hauls (fewer 
than 10%) having multiple interactions recorded (e.g., a 
killer whale or group of killer whales may be classified 
as feeding on discards as well as having the vessel crew 
use a deterrence toward it). The number of interactions in 
these tables is not adjusted for sampling rate. The mean 
between-year percentages of hauls with interactions were 
computed as the simple average between years for a speci- 
fied set of covariates. For example, the mean percentage of 
hauls with interactions for each month was computed as 
the mean between years of the percentage of interactions 
for each month. Because we are interested in comparisons 
between months of the expected monthly percentage of 
hauls where interactions occur, we use the between-year 
mean, weighting monthly estimates for each year equally. 
Alternatively, use of a weighted mean would support infer- 
ences about the probability of an interaction with a mam- 
mal being recorded for a haul in a given month, but that 
is not our aim here. 
Photographic collection and analyses 
Beginning in 1986, observers were instructed to take 
photographs of killer whales that were sighted either 
close to fishing vessels or actively interacting with fish- 
ing operations. We analyzed all photographic materi- 
als (in all forms: film, digital, and video) collected from 
1986 through 2015 to determine the killer whale eco- 
type involved in the fishery interactions that occurred in 
Alaska. A narrative was completed for each interaction, 
and the narratives include the observer’s cruise number, 
haul number, observer’s name, date, location, interaction 
