Diving behavior of air-breathing vertebrates, allometry, classification, and interspecific comparisons

Abstract

A great deal of knowledge has been acquired concerning the diving behaviour of various air-breathing vertebrates, but little has been accomplished to incorporate all of this information into general models that will explain and predict how this group of animals interact with their environment. There are three explanations for this shortcoming: 1) although diving data have been collected on many different species, very few works have attempted to summarize these findings and all of these had a fairly limited range of species, 2) methods for analyzing large data sets of multivariate behavioral observations have not yet been clearly defined, and 3) methods and techniques for analyzing diving behavior are highly variable making interspecific comparisons difficult or impossible. In this thesis, these three problems are addressed.

Maximum dive dept and duration from 129 species of air-breathing vertebrates were examined in relation to body mass. These comparisons showed that there were strong allometric relationships between diving capacity and size and that alcids, penguins, and phocid seals are all exceptional divers relative to their masses while mysticete cetaceans dive to shallower depths and for shorter periods than would be predicted from their size. Cetaceans, as well as some other groups, are probably mostly greatly affected by their feeding ecology rather than by their physiological limitations.

Next, to address the problem of how to analyze large behavioral data sets, a series of techniques for analyzing these sorts of data were tested and compared. These included k-means and fuzzy c-means clustering techniques from the field of statistics, and Kohonene self-organizing map (SOM) and fuzzy adaptive resonance theory (ART) from the field of artificial neural networks. A series of simulations were performed in order to test the performance of these techniques under various conditions. As well, real data from several species were classified to further assess the suitability of the various techniques. K-means, fuzzy c-means, and SOM all performed equally well on the artificially generated data while fuzzy ART had error rates that were twice as high. When clustering the real data, only k-means classified observations into groups that appeared biologically valid and consequently was determined to be best suited to analyze diving behavior.

Lastly, using quantitative analyses, dive data from 12 species of air-breathing vertebrates were classified, using the same technique and protocol, and interspecific comparisons were made. The behavior was classified into a series of dive shapes (depth versus time: square, V, skewed-right, and skewed-left) with a shape fitting algorithm and possible functions of the dive shapes were proposed. These were pelagic foraging, benthic foraging, exploration, travelling, resting, and food and waste processing. The observed dive patterns varied across species with body size, ecological feeding niche, and conditions of the studies from which the data came. Larger species dived deeper and longer than smaller species, as was expected, although there were a few exceptions. Also, benthic foragers usually had more square dives as well as higher proportions of bottom time during these dives than pelagic foragers. Despite these differences, strong similarities in dive shapes, the abundance of square dives, and the proportion of bottom time during square dives suggest that diving animals exploit the aquatic environment in a similar way.