Robust and realistic power analysis for cost-effective monitoring of occupancy trends

Kari E Norman, Jamie Sanderlin, Benjamin R Goldstein, Connor M Wood, Rebecca Kirby, Shaula Hedwall, Gavin M Jones

Structured monitoring programs play a key role in identifying whether a population is declining, stable, or increasing, and power analyses allow program planners to determine the number of sampling units required to reliably estimate a specified trend (e.g., 1% annual decline) with a specified error rate (e.g., α=0.20, or 80% power). Increasingly, monitoring programs track trends in site occupancy (i.e., the proportion of sampling units occupied by the target species) using dynamic occupancy models, but practical and adequate guidance for conducting power analysis using this framework is extremely limited. Here, we explored the sensitivity of sample size recommendations from Bayesian occupancy-based power analysis to variation in trend generation approach (using equilibrium, recursive, and constant survival methods), power check criteria (ranging from inclusive to strict), and a range of biological scenarios (i.e., differences in initial occupancy, survival, and detectability) and explored how these technical choices would translate to cost-optimized monitoring designs. We used a proposed real-world acoustic monitoring program aimed at tracking the recovery of the federally threatened Mexican spotted owl (Strix occidentalis lucida) as our motivating example, which requires that a monitoring program can detect a 25% occupancy decline over a 10-year period with 90% power. We demonstrated that power to detect occupancy trends is highly sensitive to trend generation approach and power check criteria. In the proposed Mexican spotted owl monitoring program, differences in trend-generation approach alone resulted in a difference in the number of sampling units required to achieve sufficient power by up to 72% (n=695 for the constant survival approach; n=1200 for the equilibrium approach) for one biological scenario, which translates to nearly $500,000 in initial acoustic equipment costs. Our study offers practical advice to practitioners interested in developing occupancy-based power analyses that allow them to robustly track realistic population changes in a cost-effective manner.