Optimal adaptive monitoring of species occupancy

Optimal adaptive sampling incorporates prior knowledge of an entity to be monitored and determines an efficient sampling design for the next sampling period based on the goals of monitoring such as maximizing the precision of an estimate, maximizing spatial coverage, identifying a best model, minimizing costs, or a balance of these.  It parallels and may be incorporated into adaptive management because the goal of monitoring may change over time.  Occupancy models are commonly used to monitor wildlife populations because they can relate landscape characteristics to species distributions, they are relatively inexpensive compared to mark-recapture, and they account for detection, yielding confidence intervals around estimates.  We will share new results for optimally and adaptively monitoring occupancy.  First, assuming all visits to all sites are of equal cost and the goal is to minimize the variance of occupancy, we demonstrate that the optimal number of visits for a site varies with both detection and occupancy.  We show through simulation that using information collected during an initial sampling period allows us create a sample design for the next sampling period with lower variance than either of the two most commonly used designs (even sampling of all sites and removal sampling) for the same cost.   We demonstrate that a lower variance (i.e. more precise estimate) could have been achieved with a lizard dataset that had 30 occasions per site by sampling each site an optimal number of times.  Finally, we illustrate how costs varying by site and visit may be incorporated to identify a sampling design that is optimal for a given budget.