Spatial modelling from line transect data

Résumé

In this paper, two new methods are presented that enable spatial models to be fitted from line transect data. Building on preliminary work by Cumberworth et al. (1996) and Hedley et al. (1997), the first method is based on a count model and involves dividing the survey effort into small segments then modelling the number of schools in each segment. In contrast, the second method uses a model based on the intervals between detections. Its formulation is derived in detail to obtain the likelihood function for the distances between detections, conditional on an estimated detection function. Both models can be fitted using standard statistical software, although variances must be estimated using computer intensive methods. We apply the methods to data from the 1992/93 IWC/IDCR Antarctic survey of Area III, fitting generalised additive models to obtain estimates of minke whale abundance, using the parametric bootstrap to estimate variance. The results from fitting these models are compared with the results of a previous analysis by Borchers and Cameron (1995}, which used conventional stratified methods.