Bias and Precision of Lizard Occupancy Estimates Vary among Observers and between Methods

Abstract

Heterogeneity in detection probabilities arises from many parts of a survey design and can introduce bias into estimates of site occupancy probability. We examined the effect of two important sources of heterogeneity, observer and survey method, on occupancy estimates in a lizard community in west Texas, with a special emphasis on Dunes Sagebrush Lizard (Sceloporus arenicolus). Using line transects and visual encounter surveys, nine observers surveyed 72 sites. We detected 10 species of lizards in total; all 10 were detected using visual encounter surveys and only 4 were detected using transects. We determined observers to be an important variable in occupancy models for two of three species of lizards. Insufficient detections were made of the remaining seven species to reliably model occupancy. We conducted simulations to further investigate the impacts of observer variability on occupancy estimate bias and precision by using a range of known occupancy and detectability combinations. Precision in the occupancy estimate increased as the mean observer detection probability increased. Overall, bias was minimized by conducting surveys with either 1) observers with high detection probabilities or 2) using teams of observers with mixed (high and low) detection probabilities.Modeling observer detection probabilities explicitly provided marginal improvement to the occupancy estimate in the simulations. The consistent variability in detection probability among observers has implications for the precision and bias in occupancy estimates and decisions should be made at the experimental design phase about how to approach this problem.