Abstract
The joint use of phylogenetic trees and ecological data has proven useful for many aspects of ecology. However, there are a multitude of phylo-diversity metrics with complex interdependencies and mathematical redundancies (the so-called ‘jungle’ of metrics). Several recentpapers have been trying to ‘map’ this jungle but appear at a first glance to contradict each other. We suggest that these contradictory results are in fact complementary and reflect two approaches to understand diversity metrics: the first focuses on general mathematical properties,the second focuses on assessing metric performance in relation to particular questions.In this manuscript, we discuss the complementarity of the two approaches and in particular how recent papers fit into this categorisation.
Main text
The joint use of phylogenetic trees and ecological data has proven useful for understanding the assembly of local communities (Webb et al., 2002; Bryant et al., 2008), for exploringlarge scale diversity patterns (Graham & Fine, 2008), and evenfor developing target conservation priorities (Isaac et al., 2007). However, the last decades have seen a proliferation of phylo-diversity metrics; we counted more than 70 (Tucker et al. 2016). Winter et al. (2013) refer to the ever-increasing portfolio of phylo-diversity indices as a “jungle”, alluding to both the multitude of metricsand their complex interdependencies and mathematical redundancies. Several groups (e.g. Vellend et al. 2010, Pavoine and Bonsall 2011, Pearse et al. 2014, Tucker et al. 2016) have tried to navigate through this jungle by exploring metricrelationships, complementarity and utility. In a recent paper,Miller et al. (2016) contribute an important discipline-specific perspective that further resolves the emerging map.
Metrics can be analysed in two ways: (1) by grouping them based on their underlying properties (e.g. by comparing mathematical formulations); and (2) by assessing context-dependentbehaviour (e.g. by comparing metric performance in relation to particular questions). The first approach requires theoretical and cross-disciplinary studies to summarize the main dimensions along which phylo-diversity metrics vary, while the second provides a field-specific perspective to quantify the ability of a particular metric to test a particular hypothesis. These two approaches have different aims, and their results are not a priori expected to be identical.
Miller et al. (2016) carry out this second approachwithin the discipline of community ecology by testing the ability of 32 phylo-diversity metrics and nine null models in discriminating between two ecological processes: habitat filtering and competition (see e.g. Hardy 2008).
The authors first simulated communities under three main assembly rules: competitive exclusion leading to species being less related than expected by chance, habitat filtering leading to species being more closely related than expected by chance, and neutral assembly. Theythen tested a posteriori which combination of metrics and null models yielded the best statistical performance. Surprisingly, only a fraction of phylo-diversity metrics and null models exhibited the ideal statistical properties of high statistical power coupled with low Type I error rate, leading Miller et al. to conclude that some metrics and null models proposed in the literature should be avoided when asking if filtering and competitionplay an important role in structuring communities. This is an important finding.
But before theoreticians run off to find new metrics, more multidisciplinary work is in order. One reason there are so many metrics is that they have been pooled across community ecology, macroecology and conservation biology. The questions typically asked by conservationists and macroecologists, for example, differ from those of community ecologists. Different metrics might perform better or worse for different types of problems. One solution would be to explicitly simulate the processes of interest for a given research question (e.g. vicariance or diversification processes in macroecological research), and select the most appropriate metric for the task. The R package presented by Miller et al., as well as others (e.g. Pearse et al. 2015) help facilitate this approach. Of course, applying all possible metrics to the appropriate simulations and null models for a given hypothesis, is complex, time consuming and inefficient. To the extent this is true, this motivates the other approach to navigating the jungle, the unified framework.
We (Tucker et al., 2016) recently took the first approach to metric analysis and classified the 70 phylo-diversity metrics along three broad dimensions: richness, divergence and regularity the sum, mean and variance of phylogenetic distances among species of assemblages, respectively-. Building upon a previous phylo-diversity classification system (Pavoine & Bonsall, 2011), which itself is based on a system forclassifying taxonomic and functional diversity metrics (e.g. Ricotta 2007, Villéger et al. 2008), the Tucker et al. (2016) classification offers clear theoretical linkages between phylogenetic and functional approaches in ecology. We then used simulations to corroborate this classification system. Although they conclude differently, we feel that Miller et al. actually provide independent support of this tri-partite classification system. Indeed, the vast majority of metrics used by Miller et al. on their simulated communities group according to this richness-divergence-regularity classification system (see Figure 1 of Miller et al.). And metrics like HAED and EED, which stem from a mathematical combination of richness and regularity dimensions, are expected to sometimes cluster with richness (as observed by Miller et al.), and sometimes with regularity (see the specific discussion on these hybrid metrics in Tucker et al., 2016).
So, while the synthesis presented in Tucker et al. (2016) takes a broad perspective across fields and across most metrics by providing an objective conceptual classification system, more focussed analyses, such as that by Miller et al., offer a detailed description of metric performance relevant to a given biological question. Both approaches have utility, and importantly, both approaches benefit each other. On onehand, detailed analyses of metric performance offer a valuable test of the broader classification system, using alternative simulations and codes. On the other hand, broad syntheses offer a conceptual framework within which results of more focussed analyses may be interpreted. For example, Miller et al. find that, in some situations, the metrics with the best statistical performances are Rao’s quadratic entropy and IntraMPD. These are hybrid richness plus divergence metrics. And indeed, the finding that metrics closely aligned with only a single dimension (called ‘anchor’ metrics in Tucker et al. 2016) do not act as the best indicators of community assembly algorithms, provides strong support for the idea that the community assembly processes modelled by Miller et al. actually involve multiple phylo-diversity dimensions. This is a characteristic that has also been recognized in the functional trait literature (see, e.g. Botta-Dukát and Czúcz 2016).
In summary, Miller et al.’s community ecology study offers an excellent complement to broad-scale syntheses such as that provided by Tucker et al. (2016). We call on researchers to continue to hack away at the jungle of phylo-diversity metrics in their own fields, in the hopes that the combination of in-depth understanding of what the metrics are (e.g., how they capture richness, divergence, regularity and combinations thereof), and how they perform under particular ecological and evolutionary processes will allow us a clearer view of our respective fields. Machetes up!