Existing and Suggested Terminology

Although the probability of resighting a bird if it is alive and present on the study area is largely a function of researcher effort, the resighting probability resulting from even the most rigorous of efforts still cannot be assumed to be 1, and will likely vary across species, sexes, habitats, and years (Martin et al. 1995). Many capture-recapture software programs exist for modeling “survival” rates and should be used in all cases so that estimates are not biased when capture probability is < 1. If sufficient data exist, a fairly complex model with separate parameters for sex, age, habitat, or other independent variables can be supported (e.g., Clobert et al. 1988, Lebreton et al. 1992).

However, an implicit assumption of these models is that the resulting “survival” estimate includes the complement of mortality and permanent emigration (Lebreton et al. 1992). Recognizing that dispersal is an inherent aspect of the ecology of these birds, even “survival” estimates that take into account resighting probability will underestimate true survival if birds disperse beyond the range of the resighting effort. While it is true that “CJS models yield unbiased estimates of survival,” it is potentially confusing with regards to site-specific resighting studies, because the estimates are still underestimates of true survival (the complement of actual mortality) if permanent emigration occurs. Even though these estimates are preferable to simple return rates, we cannot estimate true survival from local resighting studies until we can estimate the probability of dispersing “off-plot” and also the probability of settling “back on the plot” in a subsequent year (Figure 1).

We suggest alternate terminology to be used in local demographic studies that should accomplish two tasks. First, the term itself will imply what is actually measured, and second, will modify existing terminology only slightly for clarification.

Apparent Return Rate—Ratio of marked birds that are resighted on the study plot in year t + 1 to the number banded (i.e., the number that had bands) in year t.

Resighting Probability—Probability of resighting a marked bird given that it is alive and present on the study plot.

Return Rate—Probability model-based estimate of the number of birds that return to the study plot in year t + 1 to the number banded in year t.

What historically has been referred to as Return Rate is now referred to as Apparent Return Rate to imply that this measure reflects only the birds that the field researcher actually encounters. Return Rate should now be the result of a model-based estimate that adjusts the Apparent Return Rate by the probability of resighting the marked birds. Here again, we are referring to the rate at which marked birds return to the study plot. This term is equivalent to a “CJS estimates of survival,” but more clearly implies that we are not estimating true survival, the complement of mortality. To use the term “survival” in the context of local, resighting studies, we must take into account dispersal of marked birds out of and into the study plot (Figure 1).

IMPLICATIONS FOR DEMOGRAPHIC ANALYSES

Questions involving sex-specific survival, the costs of reproduction, and other aspects of life-history theory require that true survival be estimated. Once the means for separating emigration from mortality are incorporated into sampling designs that measure return rate as a response variable, modeling true survival can be accomplished. Similar model structures already exist for estimating true survival from other sampling designs. These estimates require that a portion of the birds that emigrate from the study plots be captured, resighted, or recovered elsewhere (e.g., Lebreton and North 1993 and references therein). For example, models combining the use of capture-recapture and band recovery data (Burnham 1993) seem promising as more birds are banded nationwide. Multi-state capture-recapture models (Nichols et al. 1993), in which capture-recapture/resighting data are collected at a number of isolated breeding sites, also allow estimation of true survival as well as movement rates (e.g., Spendelow et al. 1995). The MAPS program (Desante et al. 1995), which uses a constant effort mistnetting approach, is also a means by which true survival estimates can be collected by combining data from a number of trapping locations.

If the objective of the study is to compare demographic rates among different treatments, management options, habitats, or other factors, then return rate (modified by capture probability) is a meaningful parameter, because it reflects the multiple ways in which the population may be affected by the factor in question. Hypotheses regarding the effects of a particular treatment still can be tested with meaningful results, as long as it is recognized that the response variable includes both mortality and movements from the study area. One potential problem still exists regarding birds that may temporarily (i.e., for one breeding season) disperse from the study area (Figure 1) in response to a short-term treatment effect. These movements are important to detect, but currently are indistinguishable from birds that were simply "missed" during that year (Table 1).

Prompt management decisions often are needed without a complete knowledge of the system, and it may seem that underestimates of true survival are "better" because management recommendations would be more conservative. However, underestimates of survival may result in the misidentification of potential source populations (or at least stable populations) based on the measured reproductive output. Fecundity and/or juvenile survival would have to be greater than what is truly needed to balance an underestimate of adult survival. The problem is twofold, especially when considering that population models are typically female only (May and Robinson 1985) and the potential to disperse from the study is greater for females (Marshall et al. in review). First, birds that are alive, but have dispersed from the study plot, are considered dead in the "survival" component of the equation. Second, these birds are not only alive but breeding elsewhere, and their contributions to reproductive success are not included in the fecundity component. Thus we need to either incorporate dispersal from the study plot into the "survival' component, or to add their reproductive contributions to the "fecundity" component. Misidentifying a source population that is sustaining one or several sink populations could have serious conservation implications (Pulliam 1988). Pragmatic conservation recommendations with respect to these sources of uncertainty are discussed in Robinson and Morse (this volume) and Sherry and Holmes (this volume).

This cautionary note reemphasizes a longstanding recommendation to obtain better information on dispersal probabilities and distances dispersed across species, sex, and age classes for a better understanding of the population dynamics and conservation needs of long-distance migrant landbirds. Eventually, some of these problems may be solved when radiotelemetry technology allows investigators to closely follow small birds over great distances and long periods of time. Until then, we hope that some of the recommendations made here will help clarify our thinking on this subject.

ACKNOWLEDGMENTS

This paper originated from a term project conducted in RJC’s population ecology class and was enhanced by discussions with fellow graduate students and colleagues at the University of Memphis, University of Georgia, and at the workshop on ecology of migratory birds held at the University of Southern Mississippi. The manuscript was greatly improved by reviews from M. Baltz, J. Brawn, M. Conroy, C. Francis, T. Martin, C. Moore, F. R. Moore, J. Nichols, L. Powell and T. Sherry. J. Nichols and M. Conroy, in particular, clarified our thinking on this subject.

 LITERATURE CITED