Abstract
Occupancy modeling is becoming increasingly popular in wildlife management as a method of monitoring trends in wildlife populations. One of the primary motivations for the use of occupancy modeling is the ability to make inferences about large landscape patches with a reduced number of surveys. However, this increased versatility comes at the risk of l. Previous research (Hubbard 2014) has explored the presence of inherent identifiability issues in occupancy models, little work has been done on the estimability the key parameters of these models: detection probability (p) and occupancy probability (psi).
Using maximum likelihood estimation, a combination of bootstrapped profile likelihoods, data simulation and the data cloning techniques of Lele et al. 2010 were used to diagnose estimability issues across a spectrum of parameter values for p and psi, sites and surveys. Preliminary results suggest estimability issues are present at smaller sample sizes or fewer repeat surveys as either p or psi approaches the boundaries (0 or 1). The potential use of Bayesian methods to mitigate these issues is still under exploration.