Reliable population estimates (i.e., unbiased and accurate) are fundamental to effective conservation. Although raw count data is adequate to estimate population size in some instances, accurate population estimates for low-density or threatened species may require accounting for imperfect detection. Occupancy and binomial mixture models (OBMs) estimate population size from point-count data while adjusting for probability of detection. A core, but untested, assumption of OBMs, however, is that the probability of detecting a given individual within a population is independent of population density. Density-dependent behaviors linked to rates of detection may create variation in detectability among populations of varying densities. One possible consequence of a density–detectability correlation is biased estimates of population size. We tested the null hypothesis that the probability of detecting a given individual within a population is independent of population density for the endangered golden-cheeked warbler (GCWA), Dendroica chrysoparia.
Results/Conclusions
During the 2009 breeding season, we recorded the singing behavior of male warblers by means of autonomous recording units (ARU) at each of six study sites estimated to have markedly differing densities of territorial males based on long- term studies conducted within the Balcones Canyonlands Preserve, Austin, Texas (USA). We programmed the ARUs to record from sunrise to sunset for two consecutive days at each of 14 randomly selected stations within each of the six 100ha study sites. We indexed average song rate as the number of songs per unit time per study site and, using both OBMs and spot-mapping estimated warbler density per site. Analysis completed to date of a high- and low-density site has revealed that song rate per recording station within the low-density site is significantly lower than within the high-density site. When song rates are translated into rates of singing per individual male, song rate within the low-density site remains significantly lower than within the high-density site. This result is suggestive of a density–detectability correlation in GCWAs. Data analysis, when complete, will allow us to test the null hypothesis across the full spectrum of estimated densities. If a positive density–detectability correlation exist, survey protocols appropriate for calculating probabilities of detection at high-densities may result in underestimates of detectability and, thus, inflated estimates of abundance at low-density sites. The overall results of this study will illustrate the nature and magnitude of this problem for the many studies of population size in other passerine birds.