Modeling Density and Group Size of White-Tailed Deer (Odocoileus virginianus) with Distance Sampling
Management of white-tailed deer (Odocoileus virginianus) remains a high priority for wildlife agencies, and principally sound and current science is a foundation of effective wildlife management. To ensure that deer management in Indiana is grounded in sound and current science, my dissertation aims to: (1) improve components of density estimators for deer management; (2) compare candidate deer density estimators; (3) examine how spatially explicit density of demographic classes of deer change across Indiana and test for differences between demographic ratios of density; and (4) examine how deer group sizes change across space and time in Indiana. I accomplish these goals with data from fecal-pellet, camera-trap, and aerial sampling collected in Deer Regional Management Units 3, 4, and 9 within Indiana.
Two issues when estimating persistence time, , of dung piles frequently occur in deer management: (i) differences between observers on what constitutes a dung pile; and (ii) substituting the number of days between the date in which 98% of deciduous trees shed leaves in autumn and field sampling for . I therefore developed and implemented a new method for estimating , which produced density estimates that were larger than previous leaf-off methods and accounted for variation attributable to interobserver classification discrepancies. Similarly, density estimates from aerial sampling often suffers from sources of error. I showcased the importance of accounting for common types of error in aerial sampling by using a simple double-observer approach with infrared and visible cameras. My results stressed the significance of pairing red-green-blue sensors with infrared thermal sensors, choosing appropriate sampling altitudes, and using specific criteria to classify thermal signatures.
To aid decision making, I then extended cost-effectiveness analysis to choose between density-estimation methods, and simultaneously integrated precision and per-area cost of sampling, allowed for situation weighting of factors, and annualized capital cost across a single or multiple applications of capital equipment. I found aerial sampling to be the most cost-effective method for long-term deer monitoring in Indiana.
I next developed a density surface model that utilized camera-trap distance sampling within a hierarchical generalized additive model to estimate spatially explicit densities of bucks, does, and fawns. I found that deer density was influenced by landscape fragmentation, wetlands, and anthropogenic development. By extending simple statistical theory to test for differences in two ratios of density, I found strong evidence that recruitment was tied to agriculture.
Finally, I used camera traps, detectability estimates from distance sampling, and hierarchical Bayesian modelling to index group size and test multiple group-formation hypotheses in deer. I found a strong relationship between group size and several interactive predictors. I documented the largest groups in areas near anthropogenic development, in areas with high predator use intensity, and during times of day when predators were active. Additionally, groups were larger in locations with concealment when the area of concealment within the surrounding landscape was small, and larger in open areas when the amount of concealment within the surrounding landscape was large.
I lastly concluded my dissertation by encouraging future deer management in Indiana to carefully consider their goals for population estimation, and recognize and address sources of bias in common sampling protocols for population data.