The PD Phylogenetic Diversity Framework: Linking Evolutionary History to Feature Diversity for Biodiversity Conservation
Daniel P. Faith
Abstract Feature diversity refers to the relative number of different features represented among species or other taxa. As a storehouse of possible future benefits to people, it is an important focus for biodiversity conservation. The PD phylogenetic diversity measure provides a way to measure biodiversity at the level of features. PD assumes an evolutionary model in which shared features are explained by shared ancestry. This avoids philosophical and practical weaknesses of the conventional interpretation of biodiversity as based on some measure of pair-wise differences among taxa. The link to features also provides a family of PD-based calculations that can be interpreted as if we are counting-up features of taxa. The range of feature diversity calculations assists comparisons of methods, and helps overcome the current lack of review and synthesis of the variety of proposed methods for integrating evolutionary history into biodiversity conservation. One family of popular indices is based on the evolutionary distinctiveness (ED) measure. These indices all have the limitation that complementarity, reflecting degree of phylogenetic overlap among taxa, is not properly taken into account. Related indices provide priorities or other scores for geographic areas, but do not effectively combine complementarity, probabilities of extinction, and measures of restricted-range. PD-based measures can overcome these problems. Applications include the identification of key biodiversity sites of global significance for biodiversity conservation.
Keywords Option value • Endemism • IPBES • Philosophy of science • IUCN
This book addresses important concepts, methods, and applications related to the increasingly important role of evolutionary history in biodiversity conservation. The preservation of the rich heritage represented by the evolutionary history of taxa is a natural conservation goal (e.g. Mooers and Atkins 2003). This fundamental relationship between evolutionary history and conservation goals traces back at least to the IUCN 1980) proposal that taxonomically distinctive species may deserve greater conservation priority.At about the same time, Soulé (1980), in his book, Conservation biology: an evolutionary-ecological perspective, articulated a broad evolutionary perspective for conservation, and argued that “reduction of the biological diversity of the planet is the most basic issue of our time.”
The term “phylogenetic diversity” is relevant to these biodiversity conservation perspectives. The term can be traced back to the introduction of the “PD” phylogenetic diversity index (Faith 1992a, b, 1994a). PD was designed as a simple measure of the degree of representation of evolutionary history (by a given set of taxa). Faith (2002) summarised the basic definition and rationale for PD: “representation of “evolutionary history” (Faith 1994b) encompassing processes of cladogenesis and anagenesis is assumed to provide representation of the feature diversity of organisms. Specifically, the phylogenetic diversity (PD) measure estimates the relative feature diversity of any nominated set of species by the sum of the lengths of all those phylogenetic branches spanned by the set.”
That summary mentions species, but Faith (1992a, b) in fact applied PD from the outset not only to phylogenies whose tips were species, but also to phylogenetic pattern among genetic haplotypes or populations, in order to set spatial priorities to conserve within species genetic diversity (see also Faith et al. 2009). The common element across these levels is the inference of underlying diversity, where the units of variation are features or traits of taxa. This link to “features” reflects the attempt, through PD calculations, to address a fundamental concern of biodiversity conservation unknown variation, with unknown future values. Faith (1992a, b) suggested that the interpretation of phylogenetic diversity as a measure of feature diversity helps to clarify its link to conservation values: “Diversity is seen as important as the raw material for adapting to change (McNeely et al. 1990), and so provides what McNeely et al. (1990) and others call 'option value': a safety net of biological diversity for responding to unpredictable events or needs. The diversity of features represented by a subset of species provides option value in ensuring not only that one or more members of the subset can adapt to changing conditions, but also that society may be able to benefit (e.g. economically) from features of these species in response to future needs.”
Examples of these benefits include many from bioprospecting. For example, Smith and Wheeler (2006) have used phylogeny to assess potential for new discoveries of piscine venoms. Pacharawongsakda et al. (2009) have applied PD to help find natural products from microbes. Another interesting example is found in the study of Saslis-Lagoudakisa et al. (2012). Phylogenetically-related plants have provided a key medical component, discovered independently in the plants found in three different regions.
This perspective accords well with the IUCN (1980) argument for conservation of diversity in order to ensure benefits “for present and future use”. Reid and Miller (1989) echoed these ideas in their early paper, “Keeping options alive: the scientific basis for conserving biodiversity” (see also Wilson 1992; McNeely 1988; Faith 1992a, b). The Millennium Ecosystem Assessment (MA 2005) summarised this general link between biodiversity and option values: “Biodiversity loss is important in its own right because biodiversity has cultural values, because many people ascribe intrinsic value to biodiversity, and because it represents unexplored options for the future (option values)”.
Option value therefore reflects not only the unknown future benefits from known elements of biodiversity, but also the unknown benefits from unknown elements. The Millennium Ecosystem Assessment (2005) also called for “a 'calculus' of biodiversity, so that gains and losses at the level of biodiversity option values can be quantified”. These ideas are echoed in the conceptual framework for the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES; UNEP 2013) which says that values “include bequest value – in other words, the preservation of nature for future generations – or the option values of biodiversity as a reservoir of yet-to-be discovered uses from known and still unknown species and biological processes, or as a constant source, through evolutionary processes, of novel biological solutions to the challenges of a changing environment.”
The PD measure is an attempt to make inferences about “features” as units of variation, including features that are not yet known to science. Faith (1994a, b) characterised PD as one case of a general framework for biodiversity assessment that uses pattern-process models to link objects and lower-level units. In general, the biodiversity units are the things we would like to count up, and the objects contain various units. Typically, many units remain unobserved/unknown, and a patternprocess model defines relationships among the objects, enabling inference of the relative numbers of units represented by different sets of objects (Faith 1994a, b). Thus, PD provides the specific case where species (or haplotypes or populations) are the objects, features are the units, and the pattern-process inferential model is based on evolutionary processes of cladogenesis and anagenesis, manifested in phylogenetic pattern.
The link from phylogeny to feature diversity has supported the wide application of PD. For example, Huang et al. (2012) advocated the use of PD in conservation based on their finding that it provides a much stronger link to “trait diversity”, relative to species. Jono and Pavoine's (2012) study of threat diversity as a determinant of the extinction risk in mammals assessed the consequences of species declines used PD with the rationale that it “is becoming a key criterion in conservation studies because it can reflect the variety of unique or rare features of a species.”
This rationale has extended to application of PD within ecosystems, where the conservation/management goals focus on maintaining ecosystem functions and services. For example, Cadotte and Davies (2010) argued that “maximizing the preservation of PD will also tend to maximize the preservation of feature diversity, including unmeasured, but ecologically important traits” (see also Gravel et al. 2012).
Studies also link PD, feature diversity, and option values. For example, Larsen et al. (2012) argued that “it is difficult to provide a robust proxy for 'option value' – the potential value to society – as these values are not yet realized”, and concluded that “a compelling argument can be made that maximizing the retention of phylogenetic diversity (PD) should also maximize option value, as well as diversification and adaptation of the species in a future of climatic change”. The influential study of Forest et al. (2007) also highlighted the importance of PD as a link to feature diversity. They explored PD and option value based on an estimated phylogenetic tree and the geographic distribution of angiosperm genera found in the Cape hotspot of South Africa. Forest et al. (2007) concluded that, if we did not know about the medicinal, food, and other useful features of these plants, then preserving sets of species with high PD would be a good way to preserve these unknown features and their associated benefits.
PD now is regarded as “a leading measure in quantifying the biodiversity of a collection of species” (Bordewich and Semple 2012) and as “a resonant symbol of the current biodiversity crisis” (Davies and Buckley 2011), with important applications at both regional/global (e.g. Forest et al. 2007) and within-ecosystem scales (e.g. Cadotte et al. 2009). At the same time, PD must be acknowledged as just one of many biodiversity measures that are based on aspects of evolutionary history (see other chapters in this book). Unfortunately, there is no existing comprehensive review and synthesis covering all these measures. For example, Diniz Filho et al. (2013) recently concluded that “we do not even have a comprehensive and integrative approach to using phylogenies in biodiversity conservation.” Similarly, a recent review of past studies on the topic of evolutionary history and conservation (Winter et al. 2013) argued that there is little basis for distinguishing among the large number of existing phylogenetic indices (see also Devictor et al. 2010).
Partly, the existence of a gap in review and synthesis is not surprising; this area of research is evolving rapidly. The PD measure is applied in various sub-disciplines, highlighting distinctions between within-ecosystem versus global scales, microbial versus macrobial, and taxonomic levels ranging from populations to species and higher taxa (e.g., May-Collado and Agnarsson 2011; Lozupone and Knight 2005; Jono and Pavoine 2012; Jetz et al. 2014).
The other obstacle to synthesis is that, while some attempts at review and synthesis have been made, most have been incomplete or unsuccessful. Notably, philosophers of science have become keenly interested in the science of phylogeny and biodiversity conservation, but have not yet shed much light on the problem (for discussion, see Faith 2013). Philosophers so far largely have focussed on one possible unifying conceptual model of biodiversity. This model traces back to Weitzman's (1992) general framework for biodiversity, based on the idea of objects, and measures of difference between pairs of objects. The biodiversity of a given set of objects then is reflected, not in a list of the different objects, but in the amount of difference represented by the set. Weikard (2002), following Weitzman's object-differences framework, argued that “an operational concept of diversity must rely on some measure of dissimilarity between appropriately defined objects.” Maclaurin and Sterelny (2008), in their book, “What is biodiversity?”, and Morgan (2010) also saw this approach as a core framework for characterising biodiversity (the Lean and Maclaurin chapter “The Value of Phylogenetic Diversity”, also takes this as their starting point).
This approach assumes that we can decide on the definition of meaningful differences among the initial objects, and most authors have acknowledged that it is hard to choose among many possible notions of difference. This has not helped in developing a synthesis for phylogenetic measures of diversity. Winter et al. (2013) incorrectly interpreted “phylogenetic diversity” as any measure derived from a nominated between-species phylogenetic distance. Their conclusion, that there is little basis for distinguishing among different phylogenetic indices, highlighted well the problems in choosing among different notions of differences. Unfortunately, Winter et al. did not recognize PD as distinctive in avoiding arbitrary notions of difference, and instead using a model-based measure of feature diversity and option values.
A more recent study, by Kelly et al. (2014), acknowledged the feature diversity interpretation of PD, but surprisingly failed to acknowledge its pattern-process model, in which shared ancestry explains shared features. An implication of that model, emphasised from the outset, was that PD will fail to account for convergentlyderived features, and that these may be captured by an alternative pattern process model (see Faith 1992a, b, 1996, 2015). The failure to recognise these key lessons from the early work left Kelly et al. destined to merely re-discover the already wellestablished point that convergences will not be accounted for by PD, rather than making any real progress towards evaluation and synthesis (and perhaps exploring the alternative pattern-process model).
Lack of comparisons and synthesis has made it difficult to interpret some otherwise useful studies. This problem is well illustrated in the recent study by Pio et al. (2014), where “PD” is used to refer to any diversity measure linked in any way to phylogeny. They refer to a variety of published studies on the performance of “PD”, but the reader cannot know when this refers to true PD and when it refers to some other measure. Pio et al. go on to apply the actual PD method in their analyses, but without reference to that as the Faith (1992a) PD method.
Beyond the confusion in terms, there remains a genuine need to compare methods and develop synthesis. The pattern-process model approach that is the basis for PD can help in two ways. First, we can use the PD family of calculations to better recognise that there are many inter-linked, related, indices (dissimilarity, endemism, etc) rather than lots of indices that can be called “diversity” measures (for related discussion, see Sarkar 2008). In the next section, I briefly consider the PD's counting-up of features as one way to integrate other possible calculations that can be based on those counts. I then turn to the second way that PD's pattern-process model can help. Here, I will evaluate alternative measures, including those outside PD framework, by examining how well they can be interpreted under the PD features model.