Occupancy and Abundance of Golden-cheeked Warblers (Dendroica chrysoparia) on the Balcones Canyonlands Preserve

Abstract

Reliable estimates of population parameters derived from logistically feasible wildlife survey methods are essential for making management decisions regarding endangered species. Observer variability in detection can be a substantial source of ·error in avian survey methods, resulting in biased population estimates. Additionally, the degree of among-observer variability in detection may be influenced by population density. I evaluated the degree of within-and among-observer variability in detection of the federally endangered golden-cheeked warbler (GCWA, Dendroica chrsoparia) by means of point-count surveys conducted at two sites exhibiting high and low population densities. Surveys consisted of four surveyors simultaneously, but independently recording the number of GCW A detected during five-minute intervals at each of 36 points at each site. Count data were analyzed using both multi-season occupancy models and binomial mixture models (BMM) to estimate each observer's probability of detection at both the species and individual level. Model selection revealed that observer had a strong influence on detection of GCW A. I found significant variation in detection probabilities among observers and the degree of observer variability was greatest at the low density site. Extrapolating observer-specific estimates of detecting the species to four survey occasions revealed that observer variability was negligible at the high density site, yet observer variability was still substantial at the low density site. Among-observer variability in detecting individuals was more extensive at both sites, therefore I concluded that the inclusion of a covariate for observer would be necessary for modeling abundance. Herein, I also investigated the utility of point-count surveys in conjunction with occupancy and BMM as a feasible and reliable approach for monitoring the goldencheeked warbler on the Balcones Canyonlands Preserve in Travis County, Texas. Occupancy and abundance were estimated using data from point-count surveys conducted on each of five 100 hectare detection girds in 2008 and seven grids in 2009. Data were analyzed using both single season occupancy models and BMM to estimate occupancy and abundance, respectively. Occupancy estimates per grid ranged from 0.48 to 1.0 in 2008 and 0.52 to 1.0 in 2009. Estimates of abundance were compared with territory densities independently estimated using a more labor-intensive spot-mapping method. The BMM generated abundance estimates that were nearly five times as high as estimates of territory density based on spot-mapping. Thus, I concluded that BMM estimates of abundance for this species were biologically unrealistic. Using an alternative approach, I also estimated abundance using a novel C/p estimator that incorporated the probability of detecting individuals obtained from occupancy models. This alternative approach provided abundance estimates similar to territory density estimates obtained from spot-mapping. Point-count surveys conducted for this study required considerably less time and surveyed a larger area compared to spot-mapping. The results of this study suggest that using a model-based approach to estimate occupancy and abundance from point-count data is a reliable and feasible monitoring alternative to spot-mapping.