Design and analysis of quasi-experiments in landscape ecology: responses of fauna to landscape vegetation transformation in South-Eastern NSW

Abstract

The context for this work is the application of statistical science in the study of interrelationships between fauna and environments undergoing change in landscape ecology. The core of the thesis is a collection of papers which can be viewed in two ways: 1) a set of major studies in statistical practice in the field of ecology and 2) as contributions to new knowledge on the effects of landscape vegetation transformation on fauna, in particular birds, in south-eastern Australia. These papers exhibit effective contributions to ecology and conservation of wildlife through an ability to understand ecological issues from a statistical perspective. The focus in this thesis is as much on scientific process as it is on biological outcomes. The introductory chapter provides an overview of some of the statistical thinking behind my contributions to four major projects through statistical design and statistical modelling, and highlights some key findings and new understandings in ecology. Key roles in the research presented include:

• Recognising natural systems in the landscape worthy of study and seizing the opportunity to study them (ecological serendipity). ‘Treatments’ (ecological contrasts) are selected and planned and are assigned randomly in true experiments; in quasi-experiments, treatments may already exist, are not randomly assigned but can be taken advantage of. • Formulating projects and asking interesting and important questions. • Considering the ‘best’ study design to yield high quality data suitable for addressing the key questions. • Overseeing the implementation of the design, fieldwork protocols, and providing input into data collection and collation. • Formulating statistical models for analysis. • Undertaking all of the data analysis and taking responsibility for statistical analysis and resulting inferences. Where necessary, adapting existing methodologies suitable for solving the problem at hand. • Developing suitable statistical presentation of results. • Interpreting and explaining results. • Writing scientific articles and submitting them for publication.

Statistical topics and methods include the design of large-scale quasi-experiments in landscape ecology, and a range of powerful and flexible statistical modelling frameworks for the analysis of a diverse range of data types and complex data structures.

Interdisciplinary collaboration has resulted in many significant scientific contributions to ecological knowledge in the study of relationships between faunal biodiversity and landscape transformation. In particular, the importance of statistical aspects of experimental design in large-scale, long-term research studies has been demonstrated. Many new insights into effects of plantation establishment and subsequent maturation on different groups of biota have been gained. Cross-sectional and temporal responses to revegetation, regrowth, and remnant vegetation at multiple spatial scales have been quantified. Composite indices have been developed for classifying biodiversity values of farms by novel applications of statistical ideas. Furthermore, significant contributions have been made in developing sound statistical methodology for the study of long-term trends in reporting rates of birds, using structurally complex presence-absence data. These methods have been recently adapted by Birdlife Australia and now form the basis for summarising trends in terrestrial birds.