in tandem on one side of an aircraft. Their counts of S 1 =7, S 2 =3, and B=10 yielded
P 1 =0.77 and P 2 =0.59, the population estimate being Y=22 emu groups on the
843 km^2 of transects that they surveyed together, a density of 0.03 groups / km^2.
This method of simultaneous but uncolluded tallying carries two dangers, one tech-
nical, the other statistical. The two observers must not unconsciously cue each other
to the presence of animals in their field of view and ideally should be screened from
each other. Second, the chances of “marking” and “recapturing” an entity should be
uncorrelated, but they are not because marking and recapturing occur at the same
instant, the search images transmitted to each observer being nearly identical.
Caughley and Grice (1982) showed by simulation that the effect of the close corre-
lation was to underestimate density but that the underestimation became serious only
when the mean of P 1 and P 2 was less than 0.5.
Known-to-be-alive
Most estimates of population size require that the manager makes a leap of faith.
There is seldom any certainty that the population fits the assumptions of the model,
nor whether the estimate is wildly inaccurate, nor whether the confidence limits have
much to do with reality. The more complex is the model the greater is the uncer-
tainty. Many ecologists, particularly those working on small mammals, have decided
that the work needed to achieve an unbiased estimate is not worth the effort. They
would prefer an estimate which, although perhaps inaccurate, is inaccurate in a pre-
dictable direction and which does not depend on a set of assumptions of dubious
reliability. Hence the known-to-be-aliveestimate, the number of animals that the
researcher knows with certainty to be in the study area. These estimates for small
mammal populations are usually made by trapping an area at high intensity over a
short period. Each animal is marked at first capture, the estimated population size
being simply the number of first captures. Such estimates are acknowledged as under-
estimates but they have the advantage of yielding a real number, not an abstract con-
cept, to work with.
Known-to-be-alive estimates are often the most appropriate in wildlife management.
There are several problems of conservation and of harvesting for which an overesti-
mate of density may lead to inappropriate management action. An underestimate, on
the other hand, may simply produce inefficient but entirely safe management. The
penalty for a poor estimate is often distributed asymmetrically around the true
population size. It is not good to overestimate the number of individuals of an
endangered species. It is not safe to apply a harvesting quota, known to be safe for
the population size you estimated, to a population that is much smaller than you
thought. Where the undesirable consequences of an overestimate are considerably
greater than those accruing from an underestimate, the known-to-be-alive number
is often the most appropriate estimate to work with.
An index of density is some attribute that changes in a predictable manner with changes
in density. It may be the density of bird nests, or the density of tracks of brown
bears, or the number of minke whales (Balaenoptera acutorostrata) seen per cruising
hour. A common index is the pellet or fecal dropping count. This is often used in
studies of deer. It was used for the endangered marsh rabbits (Sylvilagus palustris)
in Florida, where pellet counts were closely correlated with radiotelemetry estimates
(Forys and Humphrey 1997). Active burrow entrances were used for ground
COUNTING ANIMALS 241
13.7 Indices
