AUTHORS: David Delaney, Iowa State University Tyler Harms, Iowa Department of Natural Resources Stephen Dinsmore, Iowa State University
ABSTRACT: Techniques to estimate density of unmarked animals are logistically feasible and allow sampling over greater spatial extents than more intensive methods, such as mark-recapture. However, accuracy of density estimates relies on the validity of assumptions about the study system. We conducted a thermal-imaging drone survey at night to test the validity of two assumptions for conducting distance sampling on white-tailed deer (Odocoileus virginianus) in Iowa via nocturnal spotlight surveys. First, we tested whether deer are randomly distributed with respect to gravel roads, which represent line transects in our study. Second, we quantified the portion of the population that occurs in unsampleable locations (i.e., within forest) to estimate availability bias. Preliminary analyses suggest deer do not avoid gravel roads but do responsively move away from observers prior to being detected, leading to potential bias in estimates of detection probability and density. Secondly, deer increased the use of forest cover as spring vegetation green-up occurred, leading to up to 50% of the population being unavailable to sample during surveys. Each of these deviations from conventional distance sampling assumptions inform future sampling design protocols and can be analytically corrected, once quantified, to reduce bias in density estimates.
