Needles in haystacks: Estimating detection probability and occupancy of rare and cryptic snakes

The species most in need of conservation or management are often also the most difficult to monitor, because of their rarity, secretive habits, or both. To combat these challenges, presence/absence (site occupancy) models can be used to track species occupancy at landscape scales. However, quantitative knowledge of detection probability (which is almost always <1) is required to reliably estimate site occupancy. Here, we present a case study that combines detection probabilities and site occupancy modeling to monitor a notoriously secretive guild of animals, North American aquatic snakes. Specifically, we use program PRESENCE to estimate detection probability (p) and probability of site occupancy (ψ) for seven snake species in relationship to site covariates, to understand the proximate and ultimate factors that influence habitat suitability. We were able to estimate p (3-46%) and ψ (12-96%) for each species and calculate the amount of unsuccessful effort necessary to declare absence of each species with statistical confidence (5-63 visits; 150-1890 trap-nights). We documented considerable interspecific variation in p and ψ; one species (Nerodia fasciata) was widespread and highly detectable, while another (Agkistrodon piscivorus) had low detectability despite its wide distribution. Five other species were secretive, or restricted to specific habitat types, or both, illustrating that complex and sometimes counterintuitive relationships exist between capture rate and occupancy. Incorporating p and ψ is essential to the success of large-scale monitoring programs for elusive species.