Using multiple methods to assess detection probabilities of forest‐floor wildlife

Abstract

AbstractMany previous comparisons of multiple sampling methods have assumed that detection probabilities for each method are either constant or equal to one. We used 4 sampling methods to estimate detection probabilities for forest‐floor dwelling amphibians, reptiles, and small mammals. We investigated associations between seasonality and precipitation on species detection and explored sample design tradeoffs for future studies. Although we captured 25 species, we could reliably detect (detection probability >0.15) only northern short‐tailed shrews (Blarina brevicauda) and pygmy and masked shrews (Sorex spp.) using drift fences and red‐backed salamanders (Plethodon cinereus) using visual encounter surveys (VES). The use of multiple sampling methods improved detection probabilities for only red‐backed salamanders (${\hat {p}}_{{\rm VES}} $ = 0.32, 95% CI: 0.24–0.38, ${\hat {p}}_{{\rm all methods}} $ = 0.38, 95% CI: 0.32−0.44). Parameter estimates indicated detection of both shrew species was positively related to increased precipitation. Detection probabilities for pygmy and masked shrews and red‐backed salamanders were positively and negatively associated with date, respectively. Our power analysis revealed that sampling during rain events increased the power of detecting a change in sorid occupancy by ≥40% (α = 0.05). Our results demonstrate the need to incorporate species detectability when comparing the effectiveness of different trapping methodologies. Furthermore, our study highlights the utility of power analyses for exploring study design tradeoffs for research and monitoring programs. © 2011 The Wildlife Society.