Geoscience Reference
In-Depth Information
9.6.4 Repeat Surveys
Repeat surveys provide the information that enables the occupancy status
of the units to be disentangled (probabilistically) from imperfect detection.
Although it is mathematically possible to separate the two with only a sin-
gle survey (e.g., see Lele et al
.
, 2012), it does require a number of potentially
restrictive assumptions, including that occupancy and detection probabili-
ties are functions of continuous covariates. However, the resulting estimates
are not uniquely identifiable. Swapping the covariates for occupancy and
detection provides the exact same regression coefficient estimates, although
now the inferences about occupancy and detection are swapped. Therefore,
the results are extremely model dependent.
The real value of repeat surveys is not that they enable the separation of
occupancy and detection probabilities, but that they increase the probability
of the true occupancy state being observed at least once within the sam-
pling season. That is, the true occupancy state is more likely to be eventually
observed during the defined season with additional repeat surveys; hence,
results will be more robust and less model dependent. In the case with two
occupancy states (e.g., species presence and absence), MacKenzie and Royle
(2005) showed that sampling more sites might be inefficient if there is an
insufficient number of repeat surveys (see the next section).
Repeat surveys do not necessarily imply distinct revisits to each site.
MacKenzie et al. (2006) provide some detail of different ways in which
the repeat survey information can be collected, but see also the work of
MacKenzie and Royle (2005), Kendall and White (2009), and Guillera-Arroita
et al. (2010). Some options include (1) multiple observers; (2) a single observer
conducting multiple surveys in a single visit; and (3) spatial replication
(e.g., multiple small plots within a larger sampling unit). Essentially, a sur-
vey is a single opportunity during which it is possible to detect the species.
There are, however, trade-offs in terms of how the repeat surveys are con-
ducted with respect to the effective sampling season length. For example,
nominally the sampling season might be declared as a 2-week period, but if
all surveys of a site are conducted on a single day, then the effective sampling
season is also a single day, which might have important implications for how
occupancy should be interpreted. The main considerations with respect to
conducting repeat surveys is how they may have an impact on the main
assumptions of the modeling.
9.6.5 Allocation of Effort
Given the need for repeat surveys, there are clearly competing requirements
between the number of sampling units that should be surveyed versus the num-
ber of repeat surveys per unit, particularly when the total level of effort available
is fixed and limited. Although little has been done to date on allocation of effort
for applications with multiple occupancy categories, as noted previously, it has
