Multivariate Methods for Size-Dependent Detection in Conventional Line Transect Sampling

摘要

When animals occur in groups, biases in mean group size estimates bias abundance estimates using conventional line transect methods. Multivariate models for bias correction of mean group size and for detection function analysis in conventional line transect are reviewed, developed and tested. Standard bias-correction methods, based on a least squares analysis of the observed log-transformed group sizes against estimated detection probability, tend to underestimate mean group size when outliers are present. In contrast, robust regression improves mean group size and abundance estimates when the association of perpendicular distances with other covariates is linear. Parametric and nonparametric multivariate detection function models incorporated into line transect abundance estimators provide substantial improvement in estimating mean group size and abundance.