Geoscience Reference
In-Depth Information
account of the methods available for the estimation of the encounter rate stan-
dard error using replicated transect lines, see the article by Fewster et al
.
(2009)
and Chapter 8 in the Distance user's guide. If there are not enough detections
on independent replications of lines or systematic sets of lines to obtain inde-
pendent estimates of the density, then it is recommended to use bootstrapping
to estimate the variance by resampling transect lines.
4.6 Size-Biased Line Transect Surveys
Drummer and McDonald (1987) considered the use of line transect surveys
for objects when the size of an object influences the probability of detection.
For example, Drummer et al
.
(1990) considered applications to populations
of clusters of individuals rather than single individuals (for the sea otter
Enhydra lutris
) and to pellet groups of barren ground caribou (
Rangifer taran-
dus granti
). If group size influences the probability of detection of a group,
then the observed mean group size is biased and will lead to overestimation
of the density of individuals. See the work of Buckland et al. (2001, 2004) for
more possibilities about data analysis under these conditions.
4.7 Probability of Detection on the Line of Less than One
Because of observer bias and other problems, the assumption of 100% detec-
tion of items on the transect line might not be reasonable. Quang and Lanctot
(1991) extended line transect sampling theory to the case when there is less
than 100% detection on the line and detection function is unimodal. The
assumption that there is 100% probability of detection on the line is replaced
by the assumption that a line of perfect detection exists parallel to the transect
line at some unknown distance away. Application of the technique is made
to the estimation of common and pacific loon (
Gavia pacifica
and
Gavia immer
)
in the Yukon Flats National Wildlife Refuge from aerial surveys. It is difficult
to guarantee perfect detection on the inside edge of the strip because of the
speed at which the aircraft passes a loon. The assumption of perfect detection
at a certain distance from the inside edge is more reasonable. If the assump-
tion of 100% detection on the line, on the inside edge of the strip, or at an
unknown distance away is violated, then density estimates obtained by the
line transect theory are biased underestimates. In this case, the procedure is
conservative because the population size is underestimated by some amount.
One possibility is then to use double observers for some or all of the transect
sampling. In that case, items can be detected by observer A only, observer
B only, or both observers. It is then possible to estimate the probability of
