Comparison of subjective and statistical methods of dive classification using data from a time-depth recorder attached to a gray whale (Eschrichtius robustus)

Abstract

This paper presents dive data obtained from the deployment of a suction-cup attached time-depth recorder (TDR) on a gray whale off the west coast of Vancouver Island, Canada. Data are presented in the form of dive profiles. This represents the first time that dive data have been collected from a gray whale. The data were used to compare subjective classification of dive types to statistical methods of classification, and to test the ability of the statistical methods to classify dives. Each dive was analysed using maximum depth, dive duration and bottom time variables for both subjective and statistical methods to make direct comparison of results. Subjective classification suggests that the tagged animal performed five distinct dive types. Two of these dive types, termed Interventilation and Feeding, were assigned a purpose. Two statistical techniques were then used to classify dives: k-means cluster analysis and discriminant function analysis. Cluster analysis and subjective classification showed poor agreement due to the statistical technique’s inability to account for dive geometry. Discriminant function analysis proved more successful, although this technique also demonstrated some weakness in testing for dive geometry. It was concluded that while statistical analysis of dive data is useful to classify dive types in a general manner, subtle differences, which may be indicative of behavioural differences, still depend on subjective analysis for identification. Detailed analyses of the third, or depth, dimension of the marine mammal environment will be important for the development of effective management strategies, especially as whalewatching grows in popularity.