The bird was located in location A at period 2 and survived until period 3 (asso-
ciated probability, ), it did not permanently emigrate the study system (associ-
ated probability ), it stayed in location A (associated probability, ),
it survived again until period 4 (associated probability, ) and again did not
permanently emigrate ( ), it moved to location B (associated probability, ),
and it then departed the 2-location study system (associated probability ).
Estimation under such movement models can be obtained using the multistate
modeling structures of program MARK (White and Burnham 1999) and then
modifying them to reflect the specifics of radio-telemetry data. An alternative
approach would be to write the model directly into the flexible software SURVIV
(White 1983). As noted in previous sections, point estimates themselves are not
typically of primary interest. Instead, competing models of biological interest are
developed and model testing or selection procedures (e.g. see Lebreton et al. 1992;
Burnham and Anderson 2002; Williams et al. 2002) are used to discriminate
among the competitors. As with the survival models, movement parameters can
be modeled as functions of other quantities (e.g. distance between locations, ratio
of fitness indicators in the two locations, difference in management actions
between two locations; see Nichols and Kendall 1995). Surprisingly, the sort of
probabilistic movement modeling described here has not been implemented
frequently, and reports of results from previous telemetry studies directed at
movement have tended to be descriptive. As noted in the discussion of radio-
telemetry survival studies, the assumption of detection probability equal to 1 is
not always justified in telemetry studies. In such cases, the data can be modeled
using capture-resighting models (e.g. Bechet et al. 2003).
5.4.2Capture–recapture/resighting
Various methods exist for drawing inferences about movement based on capture–
recapture data. Limited inferences are possible using capture–recapture from a
single study site (e.g. see Nichols 1996; Nichols and Kaiser 1999; Bennetts et al.
2001). For example, temporary emigration can be estimated using Pollock’s
(1982) robust design (Kendall et al. 1997). The proportion of newly caught birds
that are transients can be estimated (Pradel et al. 1997), as can departure proba-
bilities and lengths of stay on migration stopover areas (e.g. Schaub et al. 2001).
Here we focus on the use of multistate capture–recapture models for studies of
multiple locations (Arnason 1973; Brownie et al. 1993; Williams et al. 2002). Birds
are marked and released on all study locations during sampling periods and recap-
tured or resighted in subsequent sampling periods, either at the location of release
or another location. Study duration will depend on objectives. Some of the early
uses of multistate models to estimate movement were based on annual sampling
1 f 4 Bf 3 A 3 ABs 3 A
f 2 A 1 2 ABs 2 A
Movement| 131