Julian and Beeson: Estimates of marine mammal, turtle, and seabird mortality 
281 
Figure 4 
Approximate location of observed incidental seabird kills in the setnet fishery during 
the period July 1990-December 1995, all species. 
only 74 (3%) were classified as 
preset notification. In the 
driftnet fishery, out of 3,338 
observations, only 29 were 
categorized as postset notifi- 
cation. Because these data are 
so unbalanced in terms of type 
of notification, further analy- 
sis involving type of notifica- 
tion was not pursued. Apart 
from the overall difficulty of 
implementing an appropriate 
sampling plan, perhaps the 
most common difficulty was 
with accurate determination 
of location. Frequently, navi- 
gational equipment would not 
be “working” or was not avail- 
able to the observer. In some 
cases, location was incorrectly 
relayed to the observer and 
under other circumstances ob- 
servers were not able to deter- 
mine location from major 
landmarks on shore owing to 
inclement weather. A second 
difficulty was that observers 
may not have been able to de- 
termine accurately the total 
number of animals or the spe- 
cies of the animals entangled 
in gillnets owing to obstruc- 
tion of view by gear or crew. 
Also, mammals may have 
dropped out of a gill net prior 
to observation. Events of this 
type were considered non- 
sampling errors and may have 
biased estimates of incidental kill. 
Estimates of total effort were also subject to error, 
although they were treated as accurate for the pur- 
pose of bycatch estimation. They contained inaccu- 
racies due to at least three nonsampling errors: 
1 Completely unrecorded effort — A typical example 
of this was when no marketable target fish were 
caught during a trip and a logbook entry was not 
made. It was not possible to determine the size of 
this error; however, given that there were three 
chances for effort to be monitored (logbooks, ob- 
server records, and landing receipts), this error 
was probably small. 
2 Absent or imprecise reporting of location — Even 
though location data were inaccurate, a record of 
effort was tallied. Thus, because estimates were 
not stratified by area in the driftnet fishery, this 
type of error had no impact on the determination 
of estimates of total effort or mortality for this 
fishery. (This type of error may be important in 
more detailed studies.) In the setnet fishery this 
type of error would not have produced an error in 
an estimate of total effort but may have produced 
errors in effort estimates for individual geographic 
strata. For purposes of mortality estimation, the 
setnet fishery was divided into four geographic 
strata, and effort without a recorded location was 
assigned, with good accuracy, to a stratum based 
on the vessel’s known activity. If no information 
on a vessel’s behavior was known, effort was di- 
vided among strata according to the proportion 
of effort in known locations. No additional error 
