Detection probabilities, occupancy and survey effort of golden-cheeked warblers (Dendroica chrysoparia) using detection-nondetection surveys

Abstract

Surveys to detect the presence and absence of endangered species may not consistently cover an area nor account for imperfect detection. We evaluated a revised detection - nondetection survey method of the federally endangered golden-cheeked warbler (GCWA, Dendroica chrysoparia). The methodology allows a surveyor to survey consistently among sites and allowed us to estimate probability of detection (the extent of imperfect detection) and occupancy (proportion of sample units occupied at a site) using newly developed mark - recapture techniques in the program PRESENCE. From this information we then determined the survey effort required to estimate occupancy with a specified precision (coefficients of variation <0.2). Three sites were selected across the breeding range of GCWA in central Texas. At each site, 28 - 36 detection stations were placed 200 meters apart. Each detection station was surveyed 9 times during the breeding season in 2 consecutive years. Surveyors stayed up to 8 minutes at detection stations recording GCWA detected by sight or sound. We built models to assess the potential influence of environmental covariates (e.g. slope, aspect, canopy cover) on detection and occupancy and possible change in occupancy and detection probabilities within a breeding season, between years, and among sites. Using information - theoretic model selection procedures we found that detection probabilities and occupancy varied among sites, between years and within a breeding season. Detection probabilities ranged from 0.19 to 0.79 and occupancy ranged from 0.52 to 1.0. These estimates were, in turn, used to determine that 9 surveys of 32 stations at a site will be needed to have estimates of occupancy with coefficients of variation of 0.2. These findings assume no further surveys are needed in that breeding year after detection of GCWA at detection stations.