Environmental Engineering Reference
In-Depth Information
proportion of individuals in the sample that are marked provides an estimate of the
proportion of the population that was originally marked. The number in the popu-
lation is then simply the number caught on the first occasion divided by the
proportion of individuals in the second sample that bear a mark. This common-
sense calculation is in fact biased at realistic sample sizes, and although this bias
can be corrected (giving rise to the well-known Lincoln-Peterson method; Seber
1982), in practice this approach is seldom used because more complex methods are
required to detect and correct for various other sources of bias.
Modelling of mark-recapture data is implemented in a variety of specialist software
(Section 2.7.1). Full treatments of mark-recapture methods for estimating abundance
can be found in Seber (1982, 1986, 1992), Pollock
et al
. (1990) and Williams
et al
.
(2002). Here we outline the key considerations for setting up a robust survey.
Precision in mark-recapture estimates of abundance depends principally on the
proportion of the population
that is captured. As the proportion captured
decreases, the confidence interval widens rapidly, approaching infinity as the pro-
portion captured approaches zero. A substantial proportion of the population
should therefore be captured, preferably at least 50%. Unfortunately, because we
do not initially know how large the population is, or how rapidly individuals
can be caught, this does not immediately help to determine how much effort will
be needed for a mark-recapture survey. However, an assessment might be made
based on a conservative guess at the minimum expected density and probability of
capture. Alternatively, the data might be monitored over an indefinite number of
capture occasions in order to assess when a sufficient proportion of the population
has been caught. In either case, it is usually necessary to have more than two cap-
ture occasions in order to achieve a reasonable overall capture rate.
If the study area is relatively small, self-contained and easily defined (e.g. an enclos-
ure or habitat island), capture effort can be spread throughout to provide an estimate
of the total population size. Because the site has well-defined boundaries, the area is
also known precisely, allowing density to be calculated easily. More usually, capture
effort must be concentrated in a small part of the wider study site, in which case the
effectively sampled area
is not clear. A minimum sampled area can be defined by
a line drawn around the outer-most capture points, but because individuals from
outside this region are also susceptible to capture, the effectively sampled area is
greater than this. This additional area can be defined as a boundary strip around the
minimum sampled area, the width of which is defined by the target species' typical
patterns of movement. Boundary strip width can be measured in one of two ways:
●
Based on the differences in density between nested sub-grids in a regular
square grid of traps (Seber 1982; Jett and Nichols 1987). This method
requires a large number of traps and a large sample size to give a reliable esti-
mate (Wilson and Anderson 1985a).
●
As half the maximum distance moved between capture locations, averaged
across individuals (Wilson and Anderson 1985b). This method is unreliable
if individuals are rarely recaptured more than once or twice.
