Environmental Engineering Reference
In-Depth Information
mean group size to give an estimate of overall density, with overall variance incorp-
orating a component of variance in group size. In the field, perpendicular distances
to group centres must be measured, and this should be done directly during the
survey rather than measuring distances to nearest individuals and correcting for
group spread later. Such adjustment of distances after the event introduces further
error, and can lead to bias if group spread varies over time and space (Buckland
et al
. 2004).
Although group counts should ideally be made during the distance survey,
this can lead to
biased group size estimation
. For example, larger groups may be
more conspicuous, leading to over-estimation, or alternatively some individuals in
more distant groups may be missed, leading to under-estimation of group size.
There are three possible ways around this problem, at either the data collection
stage, the data selection stage or the data modelling stage. During data collection,
efforts can be made to ensure that all group counts are accurate, at least up to a cer-
tain distance away. To do this, it is acceptable to leave the line or point if necessary
to get a more complete count, although any new groups seen while away from
the line should be ignored for the purposes of abundance estimation. At the data
selection stage, if it seems clear that group counts were good up to a certain dis-
tance, data beyond this distance can be discarded from the analysis (termed data
truncation). Finally, if neither of these approaches works well, it is possible to
model variation in apparent group size with distance, thereby correcting for
incomplete data to give an unbiased estimate of group size (Buckland
et al
. 2004).
Box 2.3
Distance sampling in action: line transects for duikers.
Duikers are small African forest antelopes which are very commonly hunted.
They are usually very shy, and found in dense vegetation, making them extremely
difficult to see. Newing (1994) tested a range of methods for estimating the den-
sity of several species of duiker, primarily Maxwell's duiker
Philantomba maxwelli
,
in and around Taï National Park, Côte d'Ivoire, including distance sampled line
transects. In secondary vegetation, pilot transects yielded no sightings of duikers
due to extremely dense vegetation, so distance sampling had to be abandoned
there. However, in closed canopy forest with a more open shrub layer, sightings
were frequent enough to allow density estimation. Moving as quietly as possible
at around 1 km/h, transects were carried out both by day (yielding 41 sightings
from 33 km of transects) and by night using torch-light (yielding 46 sightings
from 23 km of transects).
Distance analysis of daytime data suggested evasive behaviour (Figure 2.5),
which was confirmed by radio-tracking data, and by occasional observations by
other researchers of duikers fleeing surveyors who were entirely unaware of the
duiker's presence! In contrast, there was no evidence of evasion during the night,
when duikers tend to freeze in the torch beam rather than fleeing. In an attempt
