Aspects of the ecology of macaroni penguins (Eudyptes chrysolophus) around South Georgia

Abstract

Macaroni penguins (Eudyptes chrysolophus Brandt) feed when diving during long foraging trips away from their breeding colony. Their foraging behaviour is constrained by a number of factors including their physiology, their condition, the condition of their partner, and the condition of their chick. This thesis concerns aspects of their foraging ecology in relation to prey availability. Time Depth Recorders and other archival tags are increasing in their capacity, battery life and durability at the same time as getting cheaper and smaller and are frequently used to explore foraging behaviour. However, development and application of techniques to analyse these resulting data sets lags behind our ability to collect the data. This thesis develops new methods and uses data from 129 macaroni penguins collected over 6 years to determine patterns of foraging behaviour. In particular: 1. Appropriate identification of sex is essential to any study of foraging ecology, in particular during the breeding season. In response to an identified problem with morphometric sexing, I conduct a validation test between molecular and morphometric sex tests and make conclusions about the nature and identity of error in these tests. 2. I use summary statistics to characterise dives. This work suggests that time series techniques may provide insights that have been lacking from previous analyses. I apply time series techniques to the data to model non-independence in data sets and to compare results obtained using auto-correlation methods with findings obtained from (1) above. The time series approach allows a comparison of different temporal elements of dives, in particular correlations between features of successive dives and how correlation between dives decays with time. 3. I use Hidden Markov Models as a clustering algorithm to provide a statistically robust description of patterns in dives that may outperform the widely used concept of bouts of dives. I then use this method to determine whether such clustering of dives exists in my data. Characteristics of bouts and types of dives vary with year, the stage and sex of the individual. 4. Hidden Markov Models do not provide a direct replacement for the concept of bouts. Rather than identify bouts as summary statistics of diving activity, I take daily summaries of activity. I then use these to overcome the nonindependence of dives and to determine the relative periods of dive activity and travel or searching. Comparison between the adult and chick fledging weights revealed a strategy of investment in the chick at the expense of the adult body mass. The application of time series techniques has led to new insights about the timing of decision-making, but this thesis reveals that further advances also require positional data to be combined with temporal activity data.