Frosted flatwoods salamanders, Ambystoma cingulatum, are federally-listed pond-breeding salamanders that inhabit wetlands embedded in pine-wiregrass ecosystems. They have likely declined due to widespread destruction of longleaf pine ecosystems and habitat degradation due to fire suppression. To explicitly address the uncertainty regarding necessary management actions for flatwoods salamander recovery, we held a structured decision making (SDM) workshop to develop an adaptive management framework for habitat restoration at one of the last strongholds for A. cingulatum-- St. Marks National Wildlife Refuge. We identified the need for state-dependent decisions, and defined possible states as combinations of habitat suitability (unsuitable [U] or suitable [S]) and occupancy (occupied [O] or unoccupied [U]), resulting in 4 combinations (S/O, S/U, U/O, U/U). We defined potential management actions for each state, and used expert elicitation to generate transition probabilities between states under two management portfolios: (A) a “future status quo” of frequent growing season burns, and (B) additional actions such as head-starting and fine-scale restoration. We then used a stochastic dynamic programming approach known as a category count model to develop a preliminary optimal management policy.
Results/Conclusions
Using a simulation approach, we estimated that U/U ponds had the highest ecological damage cost ($51,350) under portfolio A (“future status quo”). The damage cost represents the projected total cost to restore a pond in a given state (e.g., U/U) to the target state (S/O). Ponds initially in the S/U state would require $48,945 to reach state S/O, while those initially in state U/O would require $31,275. Using additional management actions (portfolio B), the projected costs of restoration decreased to $19,040, $14,345, and $11,570 for ponds initially in states U/U, S/U, and U/O, respectively – a mean decrease of 65.8%. The time to reach the S/O state also decreased under portfolio B – U/U, S/U, and U/O ponds would require 79.8%, 75.2%, and 83.6% less time to reach state S/O than under portfolio A, respectively. In a 3-pond scenario, the optimal management policy was to use additional actions (portfolio B) in most cases if the confidence in portfolio B was ≥ 40%. However, it was never optimal to use additional actions on U/U ponds. In a 10-pond scenario, some additional actions (portfolio B) were optimal if confidence in portfolio B was is ≥ 20%. When portfolio B is optimal, the reduced time to reach the target state outweighs the additional per-year cost.