Do capture and survey methods influence whether marked animals are representative of unmarked animals?

Abstract

Radiocollared animals are frequently used to estimate detection probabilities in aerial surveys. The aircraft used to conduct these surveys may also have been used to aid in the capture of these animals, and recently captured animals may react to hearing or seeing the aircraft, changing their behavior in a way that calls into question the representativeness of estimated detection probabilities. We conducted a literature review and found 30 studies that used radiocollared animals to build sightability detection models; in ≥15 of these studies, the same aircraft was used to capture and survey animals. Although it is difficult to determine whether captured animals have different sighting probabilities than the rest of the target population, we used multiyear resighting data collected from moose (Alces alces) in Minnesota, elk (Cervus elaphus) in Washington, and mountain goats (Oreamnos americanus) in Washington and Alaska, to evaluate whether detection probabilities increased or decreased as a function of time since original capture. We did not detect statistically significant within-individual time trends in detection probabilities, although we could not rule out the possibility of biologically significant results because confidence intervals on effect sizes were relatively wide. The lone exception to this rule was for the cohort of radiocollared moose in Minnesota, which exhibited a slight decrease in detection probabilities over time. Differences in detection probabilities for marked and unmarked animals may not be a significant problem for sightability models, provided that the source of the variability can be captured by model covariates (e.g., heterogeneity is tied to an individual's propensity to be in heavy cover). Nonetheless, capture-related effects likely deserve greater consideration than currently afforded, and are critically important when applying simple mark–recapture abundance estimators (Lincoln–Petersen, mark–resight) that do not allow consideration of sighting covariates. © 2015 The Wildlife Society.