A NONPARAMETRIC APPROACH TO SIZE-BIASED LINE TRANSECT SAMPLING

Point and interval estimators of the density of objects that tend to aggregate are proposed for data from line transect surveys. The probability of detection of a cluster is assumed to depend both on the perpendicular distance of the cluster to the transect line, and on its size. No particular form for the bivariate detection function is assumed, except that clusters located directly on the transect line are surely detected. An estimator of mean cluster size is also proposed. The standard errors of the estimators can be estimated either by calculating sample variances, or by bootstrap resampling. Applications of the method to a field test by Mark Otto, to a shipboard survey of minke whales, and to a survey of the bobwhite are provided. Results of some Monte Carlo simulations are presented and discussed. This method is based on the theory of trigonometric Fourier series; it appears to have the same advantages and weaknesses as the well-known Fourier series density estimator of Burnham, Anderson, and Laake (1980, Wildlife Monograph 72, Supplement to Journal of Wildlife Management 44).