Dynamic occupancy modeling for conservation
The objectives of conservation and management of animal populations are typically expressed in terms of desired values of state variables characterizing population well-being (e.g., population size; occupancy [proportion of area occupied by species]). However, the dynamics of state variables are driven by underlying process vital rates; in the case of occupancy, rates of local colonization and extinction. Management actions intended to influence occupancy generally do so by inducing changes in one or both of these vital rates, so dynamic models provide a natural approach for projecting effects of such actions. The alternative approach of directly modeling changes in occupancy as a function of management suffers from dependence on current system occupancy and vital rates, and hence lacks generality. When system dynamics are likely to be influenced by other variables (e.g., habitat change, climate change) in addition to management actions, then models projecting the joint dynamics of species occupancy and these relevant variables provide a natural approach to simultaneously dealing with both management and environmental change. The state of the art in dynamic occupancy modeling has advanced substantially in recent years to deal with added realism in both detection (e.g., misclassification models) and ecological (e.g., neighbor effects) processes, providing additional motivation for use of these models in management.
A conservation example is provided for northern spotted owls and barred owls in the Pacific Northwest. Two-species dynamic occupancy modeling provides strong evidence of effects of barred owls on spotted owl extinction rates. Projections of spotted owl occupancy dynamics in the hypothetical absence of barred owl effects show substantially higher spotted owl occupancy. Annual numbers of barred owl removals required to produce specified levels of occupancy for the 2 species are presented, together with cost estimates. This illustrative example is based on equilibria for the managed 2-species system, but optimal state-specific management actions for each year could be identified using stochastic dynamic optimization. In sum, a conservation perspective provides strong motivation for use of dynamic occupancy models, rather than static species distribution models.