Environmental Engineering Reference
In-Depth Information
Valid density estimates using distance sampling depend on a number of
assumptions
:
●
Detection at zero distance is certain.
●
Objects are detected at their initial location.
●
Distances are measured accurately.
●
Sightings are independent of one another.
These assumptions can be relaxed under some circumstances, but this generally
requires additional information, and it is preferable to design surveys to satisfy
the assumptions whenever possible. It is therefore very important that they are
understood at all stages of a survey, from planning, through implementation to
analysis.
Detection probability at zero distance
The assumption of certain detection at zero distance is the most important, and
whenever possible, the study design should ensure that it is satisfied. This is
best achieved through
training and motivation
of surveyors to ensure that the
assumption is understood, and that the central line or point is searched tho-
roughly. However, overenthusiastic searching at the centre, at the expense of effort
further out, can lead to a very rapid decline in numbers seen with distance. This is
problematic at the analysis stage, since models fitted to such 'spiked' data are
highly sensitive to minor fluctuations in frequencies, and give poor precision.
Ideally, the detection function should have a 'shoulder', being reasonably flat for
some distance from the centre before declining (Figure 2.4). This requires a bal-
anced approach to searching, ensuring that the centre is well covered, but not
neglecting greater distances.
In cases where detection at the centre is unavoidably
uncertain
, a density
estimate can still be achieved by making an independent estimate of detection
probability at the centre. The most common way to do this is by deploying two or
more independent observers on each line and recording the incidence of cases not
seen by all observers (Borchers
et al
. 1998). For this to work, it is essential that the
observers neither disturb the target objects before the others have a chance to
observe them, nor give cues to the others when they find an object.
Detection at initial locations
The assumption that objects are detected at their initial location is often a problem
for mobile animals. Movement that is
not influenced by the observer
is not a prob-
lem if the speed of movement is less than about half that of the observers. However,
if movement is much faster, care should be taken to ensure that, as far as possible,
a snapshot of animal distribution is taken within a short space of time. Failing to
do this risks animals effectively accumulating in the counted area as they pass
through, leading to over-estimation of density. This is a particular problem for
point counts of highly mobile animals.
