Estimating Detection Probabilities From Multiple-Observer Point Counts

Abstract

Abstract Point counts are commonly used to obtain indices of bird population abundance. We present an independent-observer point-count method, a generalization of the dependent-observer approach, based on closed-population capture- recapture methods. The approach can incorporate individual covariates, such as detection distance, to account for individual differences in detection probabilities associated with measurable sources of variation. We demonstrate a negative bias in two-observer estimates by comparing abundance estimates from two- and four- observer point counts. Models incorporating data from four independent observers were capable of accounting for this bias. Modeling individual bird differences in detection probabilities produced abundance estimates 15–21% higher than models that did not account for individual differences, in four out of five data sets analyzed. Although independent-observer methods are expensive and impractical for large- scale applications, we believe they can provide important insights into the sources and degree of perception bias (i.e., probability of detecting an individual, given that it is available for detection) in avian point-count estimates. Therefore, they may be useful in a two-stage sampling framework to calibrate larger surveys based on single-observer estimates. Estimación de Probabilidades de Detección a Partir de Conteos en Puntos Hechos por Varios Observadores