Environmental Engineering Reference
In-Depth Information
Furthermore, the
probability of detection
given species presence may vary in space
and time, giving rise to spurious apparent differences in occupancy.
The problem of variable detection probability can be solved, with some extra
effort, by
visiting each site several times
. Just as capture histories provide infor-
mation on individual detection probabilities in mark-recapture studies (Section
2.3.4), so detection histories can allow site-specific detection probabilities to be
estimated. In this way, an estimate of occupancy can be obtained that is not biased
by failure to detect presence in some cases (MacKenzie
et al
. 2002; MacKenzie
et al
.
2005). To use this approach, there must be
variation in perceived occupancy
—if
sites are found to be occupied either almost always or very rarely, there will be no
power to estimate detection probability. The amount of effort required depends
primarily on the desired precision and the expected detection probability, however,
as a rule of thumb, around 100 sites visited on five occasions is a realistic goal
if detection probability is around 0.25 or above. The method assumes that occu-
pancy does not change during the survey, and
repeat visits
should therefore be
made over as short a space of time as possible. Defining sites that are large relative
to patterns of movement in the target species also helps to ensure that absence on a
given visit does not simply mean that individuals have temporarily moved away (in
which case, occupancy would reflect space use rather than abundance). Detailed
recommendations for designing presence-absence surveys are given by MacKenzie
and Royle (2005).
As well as assuming constant occupancy during the survey, the basic repeated-
visit occupancy analysis also assumes that the probability of detection given species
presence is constant across sites. In order to meet this requirement, it is important
to standardise the survey, for example, by searching with
constant effort
at each
site on each occasion. Where detection probability still varies, even with carefully
controlled survey effort, and this variation is caused by variation in
local abun-
dance
, it is possible to estimate
absolute abundance
using presence-absence data
(Royle and Nichols 2003; Royle 2004; Stanley and Richard 2005). This approach
works best when numbers at each site are modest, with at least some sites unoccu-
pied, and when the variation in numbers between sites is not great. However, while
this method can provide an estimate of the total population size, there is currently
no way to define the effective area occupied by that population in continuous habi-
tat. At the time of writing, the method is therefore useful only where sampling
units have clear boundaries, within which all individuals are susceptible to detec-
tion. Tools for fitting occupancy and abundance models to presence-absence data
are available in the software PRESENCE (Section 2.7.1).
Deciding how best to estimate abundance rests on knowing which analytical
methods are available, and balancing the strengths and weaknesses of each
(summarised in Table 2.3) with practical constraints such as species characteristics,
the working environment, existing data and the financial and labour resources
