296 Monitoring Threatened Species and Ecological Communities
The dorsal fin typically bears natural marks enabling individual animals to be
identified, thus allowing the use of mark–recapture models for population
estimation. Nonetheless, the use of natural marks for individual identification is
not straightforward; misidentification (false positives and false negatives) must be
avoided if abundance estimates are to be unbiased (Yoshizaki et al. 2009), and the
proportion of animals that do not bear sufficiently distinctive marks for positive
identification must be estimated if the whole population, and not simply the
distinctively marked proportion, is to be estimated.
The sampling design for the Darwin Harbour Monitoring Program was based
on a ‘robust design’ sampling structure (Pollock et al. 1990; Brooks and Pollock
2011) of two primary samples per year (wet and dry season samples), each
consisting of nine secondary samples at each of three sites (Darwin Harbour,
Bynoe Harbour and Shoal Bay). The purpose here is to illustrate the design
principles and models employed. A more complete description is available in
Brooks and Pollock (2015).
Each secondary sample was defined as a complete set of transects through a
site. Four boats alternated each 3 days between: (1) all four operating in Darwin
Harbour; and (2) three in Bynoe Harbour and one in Shoal Bay to align the
sampling dates on the sites as much as possible to facilitate fitting multistate
models, which assume that all sites were sampled simultaneously.
Where there were sufficient data, the multistate closed robust design (MSCRD)
model (Brownie et al. 1993; Kendall 2013) was employed for analysis of the mark–
recapture data to estimate abundance, apparent survival, and movements between
sites and temporary emigration between primary samples. This approach is an
Fig. 22.3. Estimates of dugong abundance and their standard errors (vertical lines) for the aerial surveys
conducted in 2006, 2011 and 2013 using the Pollock (closed circle) and Hagihara (open circle)
methodologies. Note the aerial survey in 2006 covered a smaller area than the 2011 and 2013 surveys.
Figure reproduced from Hagihara et al. (2016).