The measurement, dynamics, and interpretation of biological diversity

The measurement of diversity reflects the variation of types (any categorical unit) across one or multiple populations and the distribution of individuals among those types. Clearly useful, a plethora of measures have been developed across widely varying fields ranging from ecology to information theory, economics, and physics. Alarmingly, however, a review of biological literature reveals considerable confusion and disagreement, made worse by the misuse and misinterpretation of measures, conflicting results, and semantic ambiguity. In order to devise a measure of diversity that can be understood across multiple research fields, it must be theoretically well-grounded, powerful, flexible, and robust. Recently, a new framework of diversity measures was developed by Reeve et al. (2016) that satisfies these criteria. This framework of measures is novel in that alpha, beta, and gamma diversity can be assessed at not only the metacommunity level but also the underlying subcommunity level. In addition to this, information on similarity between species can be tailored to suit a particular problem without changing the measures being used. This thesis examines this framework of diversity measures and explores its utility and robustness to many aspects of diversity measurement, in particular beta diversity, the measurement of variation across communities First, the ‘rdiversity’ software package was developed in R to calculate these measures (Chapter 2). These measures are then examined in detail, in Chapter 3, by comparing results obtained from these measures to known features in three distinct case studies. The first case study showcases how each measure can be used to extract different signals from a population, by investigating the spatial and temporal biodiversity of the Barro Colorado Island (BCI) Forest dynamics plot. The next two case studies demonstrate the flexibility and utility of these measures by applying diversity-based solutions to more unusual applications, investigating: the demographic diversity of the 2001 population census of England and Wales; and the transmission of antimicrobial resistance between sympatric human and animal host populations. Chapter 4 extends on the framework to develop new methods to analyse phylogenetic beta diversity. These methods are compared to traditional measures of phylogenetic beta diversity using detailed simulations. Experiments were designed to explore how well each measure was able to detect phylogenetic signals in community structure (varying the number of tips in the phylogeny, the number of subcommunities, evolutionary rate, whether a phylogeny is ultrametric or non-ultrametric, whether data is incidence-based or abundance-based, nestedness vs. turnover, and so on). Results showed that these measures, particularly the exponentially transformed phylogenetic distance-based beta diversity measures are the most robust across all measures tested, having the greatest power to detect community structure in almost all cases across all measures tested. Following this, a case study highlights the utility of these measures, using phylogenetic data to assess transmission of antimicrobial resistance between human- and animal-origin isolates of Salmonella DT104. In the final results chapter (Chapter 5), I consider the issue of inaccuracies arising from incomplete sampling, which are ubiquitous in diversity measurement. The robustness of the diversity framework is tested comprehensively under two distinct sampling strategies, reducing the sampling effort per unit area and reducing the area sampled. This is explored using the fully sampled 50 ha BCI Forest dynamics plot dataset, where results show that subcommunity measures are particularly robust to subsampling by subcommunity (reducing the area sampled). The aim of this thesis is to demonstrate that whilst the concept of diversity continues to be shrouded in ambiguity – where a review of literature reveals as many measures of diversity as possible research questions – there now exists a single framework of flexible and robust measures capable of detailed analyses across different data types, resolutions, and applications.