282 
Fishery Bulletin 96(2), 1998 
was included in variance estimates because strata 
were large enough and vessels predictable enough 
that “assignment error” was most likely negligible 
compared with other assumptions. There was no 
way to determine accurately the direction of bias 
in effort estimates for each area. 
3 Underestimation of effort from landing receipts — 
Underestimation of effort from landing receipts 
occurred when no effort was logged by fishermen 
or recorded by a NMFS observer for three days prior 
and subsequent to the date of the landing receipt. 
In these cases, a landing receipt was assumed to 
represent one day of fishing effort. Of the three types 
of nonsampling errors examined, this was poten- 
tially significant for the driftnet fishery because trips 
in this fishery may last over two weeks. For example, 
in 1993, the median length of a driftnet trip was 7 
days; the mean was 6.8 (SE=3.3). Consequently, the 
number of times this approximation occurred was 
important. In the setnet fishery, with shorter trips, 
this assumption had less impact. It applied to mul- 
tiple-day trips around the Channel Islands area, 
off the coast north of Santa Barbara, or when a fish- 
erman held catch for more than one effort-day. 
Setnet trips were seldom more than a day in length 
because of catch storage and preservation limita- 
tions. In both the drift and setnet situations, the 
minimum effort representing a landing receipt was 
used, namely one effort-day, because a more appro- 
priate approximation would have required a case- 
by-case investigation. The assumption contributed 
toward a possible negative bias of estimates of total 
effort and incidental kill. 
Figure 5 
Approximate location of observed incidental marine mammal kills in the setnet fish- 
ery during the period July 1990-December 1995, all species. 
Combined, these three dif- 
ficulties in estimation of total 
effort may have produced a 
negative bias in estimates of 
total effort and, consequently, 
in estimates of incidental kill 
(i.e. current estimates of total 
effort may reasonably be con- 
sidered a minimum). An accu- 
rate estimate of bias would be 
difficult and costly to obtain. 
Mortality 
Infrequent entanglement of 
marine mammals, turtles, 
and seabirds makes accurate 
estimation of species-specific 
incidental kill difficult. In 
spite of our good intentions, 
estimates may have been con- 
founded by distribution of 
fishing effort or movement of 
species. In the future, the 
variability of estimates can be 
decreased by more strictly 
implementing a sampling 
plan. For some species, a 
posthoc analysis may be pro- 
ductive as information on 
movement and abundance of 
stocks becomes available. For 
infrequently entangled spe- 
cies, only increased observer 
coverage will improve estima- 
tion of incidental entangle- 
ment and kill rates. 
