Estimating Individual Detection Probability for Stream-Associated Amphibians Using Multiple Visit Surveys

Most stream-associated amphibian studies have used count data to index abundance (Kroll 2009). Inferences based on comparisons of these indices over time or space assume that detection probability remains constant. As part of an experimental study examining the effectiveness of different riparian buffer prescriptions on non-fish-bearing stream basins, we utilized recently developed models (Royle 2004) to estimate individual capture probability using data collected from in-stream amphibian surveys repeated over multiple sampling occasions during the pre-harvest interval.

These models consider site-specific population size, N, as independent random variables that are assumed to be distributed according to some mixing distribution. Thus prior parameters are estimated from the marginal likelihood of the observed count data, having been integrated over the prior distribution of N. This technique also allows for modeling of covariate effects on detection probability and abundance.

Our objectives were to determine how detection probability varied for the 3 species and if different covariates were associated with detection probability.