Employing Relative Entropy Techniques for Assessing Modifications in Animal Behavior

Enrico Scalas,Minoru Kadota,Tsutomu Takagi,Eric J White,Shinsuke Torisawa,Kazuyoshi Komeyama

doi:10.1371/journal.pone.0028241

Abstract

In order to make quantitative statements regarding behavior patterns in animals, it is important to establish whether new observations are statistically consistent with the animal's equilibrium behavior. For example, traumatic stress from the presence of a telemetry transmitter may modify the baseline behavior of an animal, which in turn can lead to a bias in results. From the perspective of information theory such a bias can be interpreted as the amount of information gained from a new measurement, relative to an existing equilibrium distribution. One important concept in information theory is the relative entropy, from which we develop a framework for quantifying time-dependent differences between new observations and equilibrium. We demonstrate the utility of the relative entropy by analyzing observed speed distributions of Pacific bluefin tuna, recorded within a 48-hour time span after capture and release. When the observed and equilibrium distributions are Gaussian, we show that the tuna's behavior is modified by traumatic stress, and that the resulting modification is dominated by the difference in central tendencies of the two distributions. Within a 95% confidence level, we find that the tuna's behavior is significantly altered for approximately 5 hours after release. Our analysis reveals a periodic fluctuation in speed corresponding to the moment just before sunrise on each day, a phenomenon related to the tuna's daily diving pattern that occurs in response to changes in ambient light.

Highlights

Understanding the movement patterns of animals is crucial for their proper management and conservation
We show how methods from information theory can be used to quantify the gain in information provided by a newly measured observation, relative to a known reference distribution
This measure, which serves as an effective distance between the observed and reference distributions, is called the relative entropy

Summary

Introduction

Understanding the movement patterns of animals is crucial for their proper management and conservation. In order to properly understand the relationship between an animal’s behavior and its environment, it is essential that researchers determine the possible effects that transmitter attachment and presence can have on equilibrium behavior and physiology. The attachment of a transmitter can induce stress in the animal, thereby interrupting its normal foraging behavior. In such cases the speed of an animal may be a good indicator of non-equilibrium behavior. Measurements of speed distributions for newly tagged animals can be significantly different from speed distributions under normal behavior. Mild stress may not always be reflected by statistically significant changes in the speed, even though an observer confidently asserts that the animal is not behaving normally

Objectives

Methods

Results

Conclusion