relatively short, should include relatively little mortality, and should occur at
a time of the year when birds are relatively stationary. The recovery period need
not be short and may include either the hunting season or the entire year.
Because birds are encountered twice at most (initial banding and possibly recov-
ery as a dead bird), it is not possible to estimate survival rates based on banding
of young (first-year) birds only, unless fairly restrictive modeling assumptions are
made (Brownie et al. 1985). Banding studies relying on recoveries, rather than
recapture or re-observation data, should thus include banding of both adults
and young birds. Additional information on study design and sample sizes is
presented by Brownie et al. (1985) and Williams et al. (2002).
5.4 Movement
5.4.1Radio-telemetry
Although satellite telemetry studies can effectively record bird locations any-
where on earth, the majority of radio-tracking studies of bird movement still
involve one or more local study areas. We consider both the field situation and
statistical modeling for one and multiple areas separately. If detection probability
for radioed birds is 1, then studies of a single area can be used to address questions
about the probability of a bird departing the study area either permanently or
temporarily. As described for radio-telemetry studies of survival, field sampling
requires periodic sampling of the study area. At each period, instrumented birds
are located and designated as either alive on the study area (denote as 1), dead
(reflecting death following the previous sample period) on the area (denote as 2),
or not present on the area (denote as 3), indicating movement off the area
following the previous sample period. A 0 is then used to denote the sample
periods before an animal is tagged and after an animal has died or departed the
study area. Thus, encounter history 0 1 1 1 3 0 would indicate a bird radio-
tagged during sample period 2, relocated on the study area in periods 3 and 4,
and not present on the area in period 5 (hence moved off the study area).
As with survival estimation, we must develop a probabilistic model to describe
the sequence of events depicted by the encounter history. In survival estimation,
the event of interest is bird death, whereas in this type of movement modeling, the
event of interest is departure from the study area. Also as with survival modeling,
two basically equivalent ways exist for conducting this modeling, one based on
the time elapsed until movement and the other based on a binomial model of
movement between each pair of sampling periods (see Bennetts et al. 2001;
Williams et al. 2002). Here we outline the binomial modeling approach. Define
sias the probability of survival from sample period itoi1 for a bird that
remains on the study area and fias fidelity, or the probability that a bird alive and
Movement| 129