No branch left behind: tracking terrestrial biodiversity from a phylogenetic completeness perspective

Abstract

Biodiversity, as we see it today, ultimately is the outcome of millions of years of evolution; however, biodiversity in its multiple dimensions is changing rapidly due to increasing human domination of Earth. Here, we present the “phylogenetic completeness” (PC) a concept and methodology that intends to safeguard Earth’s evolutionary assets that have arisen across the tree of life. We performed a global evaluation of the PC approach using data from five major terrestrial clades and compared the results to an approach in which species are conserved or lost randomly. We demonstrate that under PC, it is possible to maximize the protection of greater evolutionary assets of each clade for a given number of species extinctions. The PC approach is flexible and can be used to assess biodiversity under different conservation scenarios. The PC approach complements existing conservation efforts and is linked to the post-2020 Convention of Biodiversity targets.

Introduction

Over more than 3.5 billion years of life on Earth, evolution has generated and honed a vast array of innovations represented by the diversity of forms and genomes across the tree of life. Contemporary species collectively represent the genetic assets that contribute to the functioning of the current biosphere, and these functions in turn serve as the foundation of Nature’s Contributions to People (NCP) (Díaz et al. 2019). Put another way, species embody evolutionary innovations that represent complex and unique approaches to life on Earth. These innovations not only support current NCP, but are also necessary for future benefits to humanity, including those not yet discovered.

Phylogenetic trees depict the hierarchy of life in which species are nested in larger and larger clades, each descended from a more distant common ancestor (Figure 1). They provide information on the breadth and variation of innovations evolution has generated and can be used to inform approaches to species conservation with the goal of minimizing extinction of evolutionary innovations (Faith 2002; Larkin et al. 2016). Close relatives typically have a high proportion of shared genetics since they arose from a common ancestor at some point in the comparatively recent past, and thus share many of the same innovations.

Figure 1.

Schematic diagram of the phylogenetic relationships of species within a lineage of organisms, showing the hierarchical nesting of hypothetical species. The number of species that need to be conserved in order to prevent the extinction of any branches in the tree depends on the depth in the phylogeny we define the branch. Dashed lines represent different ages (T_N) or depths in the phylogeny used to define branches—the corresponding number of species preserved (T_SP), if one species per branch is conserved, is colored the same as the dashed line. For example, purple circles (fewer branches and species with a deep slice at T_N3) to red ones (more branches and species with a more recent slice at T_N1). The deeper we “slice” the phylogeny, the fewer the species need to be saved in order to preserve all branches of the phylogeny at that slice. Black circles represent species at T_N0, i.e., the current species, assuming no extinction. Open circles filled with “x” indicate species that went extinct at a specific T_N.

A wide range of phylogenetic metrics exist that capture diversity across the tree of life (Tucker et al. 2017) and are relevant to discerning how much variation is captured under different conservation scenarios. Most metrics of phylogenetic diversity are applied to phylogenetic trees and represent a sum over phylogenetic branch lengths (possibly after augmentation to account for other factors, such as probability of extinction). The canonical example is phylogenetic diversity (PD, sensu (Faith 1992)), which, for a subset of species, is defined as a sum of all branch lengths required to connect those species (Faith 1992, 2002).

In our era of rapid biodiversity loss (Tilman et al. 2017; Díaz et al. 2019) (Tilman et al. 2017; Díaz et al. 2019), keeping all species—the tips of the tree of life—is not realistic, and indeed is impossible given recent extinction rates (Rounsevell et al. 2020). However, it may still be possible to conserve all branches of the tree of life, depending on how deeply we define the branches. This concept is the principle behind phylogenetic completeness, a concept and methodology we propose here that aims to preserve Earth’s evolutionary assets that have arisen across the entire tree of life. Phylogenetic completeness (PC) and phylogenetic diversity (PD) both aim to maintain as many branches as possible from the tree of life, recognizing it is not possible to retain all species. But given the hierarchical nature of phylogenetic trees, how a lineage is circumscribed—or what constitutes a “branch” that should be protected—can influence the outcome. PC differs from PD in that it slices across the tree at a given point in evolutionary history and defines a set of branches based on this cut off (Figure 1), using an accounting framework, rather than merely maximizing phylogenetic breadth.

Here we conducted a series of analyses to explore the implications for conservation depending on the depth in the tree of life where a “branch” is defined. The goal was to develop and apply a framework for conservation of species that minimizes hemorrhaging of Earth’s evolutionary assets, given a fixed level of species loss that is assumed to be unavoidable. We then compared the loss of PD under conservation scenarios in which the set of species targeted for conservation was based on an “informed” phylogenetic completeness approach or in which species conserved or lost to extinction occurred randomly.

Methods

Phylogenetic completeness (PC) approach

We developed an approach in which a phylogeny is iteratively sliced at different periods of time (T_N) until a specified finish time (T_F). For example, if a phylogeny is sliced every T_N = 2 million years until T_F = 50 million years, a total of 25 slice points is obtained (see also Figure 1). These slice points are then used to drop all but one terminal tips—or operational taxonomic units (OTU)—from the phylogeny; with this approach, we ensure that at least one OTU of each lineage at a specific time (T_N) is kept. In other words, by keeping at least one OTU from each lineage in the tree of life, we aim to maximize the preservation of the deepest evolutionary history. At each slice point (T_N) we additionally calculated the number of species (T_SP) and phylogenetic diversity (T_PD) as the simple sum of branch lengths at the specific slice point (T_N).

Empirical assessment

The empirical assessment was focused on five major terrestrial clades (seed plants, amphibians, squamates, birds, and mammals). Description of the data can be found in the WebPanel 1.

We tested the reliability of the PC approach by slicing each phylogeny every T_N = 100 Ky until T_F = 100 My and calculated the T_SP and T_PD at every slice point. These metrics were then used to identify change points in the phylogenetic diversity over the 100 My time period. Change points were evaluated using Bayesian Multiple Changing Points (MCP) regressions. The first changing point plus its credible intervals (CIs) identified by the MCP analysis were used as cutoff thresholds to estimate the number and identities of OTUs to be kept. This procedure allowed us to identify different change points or cutoff thresholds in the phylogenetic diversity over 100 My for each clade separately and consequently prevent us establishing a fixed arbitrary cutoff threshold (e.g., setting a changing point at 2 My as cutoff threshold) for all clades.

We compared the diversity in each clade for a PC conservation scenario, in which species were managed to maintain all phylogenetic branches to random losses (RDM). In other words, we removed OTUs at random until the identified cutoff threshold for each clade separately. This procedure was repeated 1000 times, and at each step the T_SP and T_PD were estimated.

Finally, using the OTUs identities from both the PC and the random loss scenarios we mapped the phylogenetic diversity of seed plants and terrestrial vertebrates globally. These maps were used to estimate the difference (ΔPD) between the observed PD (PD_OBS) and the expected PD (PD_EXP) under either the phylogenetic completeness (PD_PC) and the random loss approach (PD_RDM).

These maps represent the proportional difference between the observed (PD_OBS) and expected (PD_PC or PD_RDM) phylogenetic diversity, where negative values suggest that a grid cell will lose a proportion of its PD according to a specific conservation scenario, e.g., under the PC scenario. Note that the lower CI from the Bayesian MCP regressions were used as variable cutoff thresholds for each clade for mapping purposes.

Protected areas assessment

We were also interested in evaluating the role of protected areas (PAs) in protecting the evolutionary history of terrestrial biodiversity. To do so, we overlayed the PAs with the vertebrate and seed plants PAMs to obtain the presence-absence of species within the PAs (PAM_PA) globally (WebPanel 1). These PAM_PA were then used to estimate the number of species and phylogenetic diversity within each PA. Using this information, we estimate and map the ΔPD between the observed PD (PD_OBS) and the PD within PAs (PD_PA). The resulting maps show how much evolutionary history are currently protected within the PAs.

Results

Phylogenetic completeness

Bayesian MCP models revealed variable cutoff thresholds for each clade (Figure 2; WebTable 1). Based on these thresholds, losses of species ranging from 1.34% to 18.11% are estimated to occur in each of the major clades (seed plants, amphibians, squamates, birds, and mammals) (WebTable 1) while still safeguarding between 97.27% to 99.97% of the phylogenetic diversity—i.e., of the evolutionary history of each clade (WebTable 1; Figure 2). If the lower credible interval (LCI) of our Bayesian model estimates were used to define the phylogenetic branches to be conserved (Figure 2; WebTable 1), a higher number of species and branches in the tree of life would be safeguarded (Table 1).

Figure 2.

Changes in the phylogenetic diversity and number of species over 100 million years for terrestrial biodiversity, including seed plants (A), woody plants (B), amphibians (C), squamates (D), birds (E), and mammals (F). Black dashed vertical lines represent the first changing point identified by the Bayesian Multiple Changing Points (MCP) regressions. Black dotted vertical lines represent the 95 credible intervals. Red long-dashed horizontal lines represent the observed number of species for each clade. Red dashed horizontal lines are the number of species expected under the first changing point. Red dotted horizontal lines represent the expected number of species under the lower bound (or credible interval) of the first changing point. See WebTable 1 for a comprehensive numerical summary of the changing points. Silhouettes obtained from Phylopic (http://phylopic.org).

Table 1.

Number of branches (NB) conserved under the scenarios of phylogenetic completeness (PC) and random loss (RDM) conservation scenarios. The number of branches conserved under the phylogenetic completeness approach is higher than the random loss scenario for the same number of species except for birds and mammals. Percentages of NB under PC and RDM scenarios are displayed within brackets. Number of branches for all species within a clade assuming no extinction is displayed for reference. The number of species under PC/RDM correspond to the number of species at the minimum threshold (lower credible interval) identified using the Bayesian MCP regressions. Note that the observed number of species (second column) corresponds to the number of species sampled in each phylogenetic tree and might not represent each clade’s true number of species.

These analyses demonstrate that if conservation efforts are focused on maintaining defined branches of the tree of life it is possible to maximize the accumulated evolutionary innovations that are safeguarded across all clades even when individual species go extinct. Figure 3 shows the comparison between the estimates of PD under both phylogenetic completeness (PC) and random loss (RDM) scenarios. Under PC, a higher number of branches (Table 1) and greater evolutionary history in each clade is preserved for a given number of species extinctions (Figure 3).

Figure 3.

Difference between the remaining phylogenetic diversity under phylogenetic completeness (red) and random loss (green) scenarios. X-axis was log-transformed for plotting purposes. In all cases, the phylogenetic diversity is higher for the phylogenetic completeness approach.

Spatial patterns of ΔPD under PC and RDM scenarios (Figure 4; WebFigure 1), show how conservation informed by PC safeguards a greater proportion of evolutionary history even with the same number of species extinctions. For example, for seed plants in tropical regions across the world, conservation informed by PC resulted in PD loss below 10%, whereas the RDM scenario resulted in 10-20 % of PD loss. Extinction patterns of terrestrial biodiversity under PC and RDM scenarios at the biome level (WebFigure 2) also show greater preservation of accumulated evolutionary innovations when conservation is targeted to maintain branches of the tree of life. Nevertheless, we find that Tundra and Taiga biomes are susceptible to high losses in PD, especially for seed plants, birds, and mammals, even under PC scenarios. In contrast, tropical biomes (for both forest and grasslands) show limited losses in PD for the same threshold values used to define branches, as in Tundra and Taiga biomes (WebFigure 2). These results indicate that for the a given number of species extinctions, tropical biomes will lose fewer branches of the tree of life and are thus less susceptible to loss of evolutionary history.

Figure 4.

Mapped phylogenetic diversity for seed plants (A-B) and birds (D-E) globally under scenarios of phylogenetic completeness and random loss. Legends indicate the proportional loss of phylogenetic diversity (ΔPD). Blue tones indicate that more branches of the tree of life have been preserved and red tones that more branches have been lost. Comparing the two scenarios, globally more branches of the tree of life are conserved in the phylogenetic completeness scenario for the same number of vertebrate species extinctions. Bottom panels (C and F) show the differences in phylogenetic diversity (ΔPD) in 1° grid cells between the species currently estimated to occur in those cells and the fraction of those estimated to occur in protected areas globally. Red colors represent greater ΔPD, meaning more branches of the tree of life are not currently protected. Continental China is shaded gray given that protected area information is currently unavailable in the WDPA for this country. Maps for amphibians, squamates, and mammals can be found in the supplementary material.

Protected areas assessment

The currently implemented PAs across the world cover an area of approximately 26,775,820 km 18% of the land surface. Collectively, ~97% of the species in our dataset overlap their ranges with the system of PAs globally (Table 2). Although the proportion of likely protected species is high, a large proportion of the Earth’s land surface, and thus most of the ranges of most of these species, is not protected (Figure 4; WebFigure 3) and is threatened by land use change and other human activities.

Table 2.

Summary of terrestrial biodiversity within protected areas. Table summarizes the number and percentage of species currently protected in protected areas. Our estimations of %protected and % of loss are based on the number of species available in the geographical ranges and phylogenetic datasets (see methods). Note that the number of seed plant species with available geographic ranges is 207,146, so the estimations of %protected and % of loss may be underestimated. The observed number of species (second column) corresponds to the number of species sampled in each phylogenetic tree and might not represent each clade’s true number of species.

Discussion

Our “phylogenetic completeness” framework for informing biodiversity conservation focuses on maintaining the accumulated evolutionary innovations across the tree of life with the intent of leaving no branch behind. We introduce a rigorous approach for defining branches across clades of terrestrial organisms to ascertain where in the tree of life there is high evolutionary redundancy and where a single species may represent an entire branch. In doing so, we establish phylogenetic branches as units of conservation priority. By defining these branches and the species contained within them, the phylogenetic completeness approach provides critical information on branches at risk of extinction where there is low redundancy as well as flexibility in which species can targeted for conservation in cases of high redundancy. The approach is particularly useful in developing conservation priorities in relation to protected areas by tracking which branches of the tree of life are currently safeguarded and by identifying the branches that are at highest risk—those not currently protected or with the least amount of their range protected. This approach is consistent with the Convention on Biodiversity draft Milestone A.2, that the increase in the extinction rate is halted or reversed, and the extinction risk is reduced by at least 10 per cent, with a decrease in the proportion of species that are threatened, and the abundance and distribution of populations of species is enhanced or at least maintained. Yet it also captures elements of diversity not delineated by the CBD by accounting for the breadth of evolved variation in plant and vertebrate-animal life.

Increasing human domination of Earth and its ecosystems is rapidly changing biodiversity patterns and negatively impacting the capacity of ecosystems to provide goods and services to humanity (Tilman et al. 2017; Díaz et al. 2019). Safeguarding all remaining biodiversity, although desirable, is unrealistic on the basis of virtually all projections (Pimm et al. 2014; Urban 2015; Tilman et al. 2017); the footprint of humanity is currently too large to completely avoid further extinctions. Scientists have recognized the challenge of developing logical conservation solutions given the complexity of stakeholders, managers, and indirect actors as a ‘wicked problem’ that has no straightforward solution (Vane-Wright et al. 1991; DeFries and Nagendra 2017). Focusing on the conservation of evolutionary history has been hailed as an integrative way to safeguard most of the biodiversity and its functions (Faith 1992; Mooers 2007). For example, a recent study by Molina-Venegas and collaborators (Molina-Venegas et al. 2021) (Molina-Venegas et al. 2021) found strong evidence that plant evolutionary history is tightly linked to multiple plant use categories and therefore to human well-being (Molina-Venegas et al. 2021). These findings, among others, support the idea that conserving evolutionary history is critical for future human well-being (Forest et al. 2007; Molina-Venegas et al. 2021).

Multiple approaches have been proposed to assess changes in biodiversity focusing on “hotspot” areas (spatial prioritization), or taxa (taxonomic prioritization) for conservation purposes (Margules and Sarkar 2007). These approaches rely on the use of metrics that capture different dimensions of biodiversity, e.g., metrics that capture evolutionary changes among a set of taxa (Margules and Sarkar 2007) or the variation in form and function of taxa within communities (Díaz and Cabido 2001; Petchey and Gaston 2006). Despite their usefulness for assessing the state and the fate of biodiversity, most of these metrics, if not all, are sensitive to information completeness. Missing information can result in misleading metric calculations and inappropriate interpretations of spatial or taxonomic comparisons (e.g., Weedop et al. 2019). The PC framework introduced here represents a complementary approach to counting numbers of species or comparing levels of phylogenetic diversity to assess biodiversity under alternative conservation scenarios. It provides an accounting framework that prioritizes conservation of branches of the tree of life rather than individual species (Table 1; Figure 2). It also allows the identification of areas susceptible to high losses of evolutionary assets (Figure 4; WebFigure 1), which can be used as baseline information for spatial prioritization, providing a broader context for local decision making (Chaplin-Kramer et al. 2022). Moreover, in the context of spatial prioritization, the PC framework may be less susceptible to missing data given that it focuses on the branches of the tree of life. To illustrate this point, if we consider protecting at least one descendant taxon from a specific node in the phylogeny, this taxon contains genetic information that captures most of the evolutionary history of the entire branch (see Figure 1). If species within the branch have not yet been identified or are not readily observed, the branch itself is still preserved, with the caveat that phylogenetic information remains imperfect.

Substantial efforts have been invested to prevent the extinction of biodiversity. In particular, the establishment, expansion, and enhancement of protected areas (PAs) have received considerable attention due to their critical role in protecting Earth’s biodiversity and preventing the erosion of its benefits to humans (Naughton-Treves et al. 2005; Watson et al. 2014). Despite important progress in establishing terrestrial PAs around the world (Table 2), several critical areas are still unprotected, leaving many branches of the tree of life vulnerable to extinction (Figure 4; WebFigure 3). Note that in our evaluation we used the species presence within current PAs to estimate which species are less vulnerable to extinction for the purposes of deciphering which branches of the tree of life are most at risk. We thus assume that species not present in PAs are the most vulnerable to extinction and in greatest need of conservation or assistance.

At high latitudes (tundra) an individual species frequently represents an entire phylogenetic branch while at low latitudes (tropics), a branch is likely to contain many species. This pattern is largely the consequence of more recent divergence times and higher speciation rates in the tropics. However, the spatial scale (grain size) must also be considered, for example, one hectare in a tropical forest can hold ~650 tree species more than all tree species that occur at high latitudes (Coley and Kursar 2014). Despite this high diversity, tropical forests are usually hyperdominated by a fraction of species (~1.4% of about 16,000 tree species estimated for the Amazonian Forest are considered as hyperdominant) that are specialists to their habitats and have large geographical ranges (ter Steege et al. 2013). The less abundant species or poorly known species with small geographical ranges are potentially threatened. Although it is beyond the scope of this article, PC evaluations at local and regional scales could help to identify which species may be prioritized to prevent losing branches of the tree of life at local and regional scales.

Furthermore, the combined effects of climate and land-use changes are likely to limit the role of PAs in safeguarding biodiversity (Hoffmann et al. 2019; Asamoah et al. 2021). About one-quarter of the PAs globally are projected to suffer rapid climate change, with small PAs being the most impacted (Hoffmann et al. 2019; Asamoah et al. 2021). Advancing projections of combined limits to species tolerance to climate change and migration capability without assistance could be oriented towards understanding which branches of the tree of life currently protected are most likely to be lost. Doing so, could help inform costly efforts to establish corridors for movement within human-dominated landscapes (Gibson et al. 2011; Wintle et al. 2019) and assisted migration plans to target the protection of species that maintain all the branches of the tree of life.

Conclusion

In our era of rapid global change and rapid biodiversity loss, conservation efforts must reckon with the reality that we will not succeed in saving all species on Earth. We outline an approach for conservation, which we call phylogenetic completeness, that focuses on saving the accumulated innovations that have evolved in Earth’s biota by counting individual branches in the tree of life as units of conservation priority. The approach benefits from detailed information of the tree of life that is only now sufficiently resolved to be applicable to all of life on Earth. The approach complements other conservation efforts and is directly relevant to the targets of the post-2020 Convention of Biodiversity.

Statement of Authorship

All authors contributed intellectually to the manuscript. J.C.-B. and J.N.P.-L. conceived of and framed the manuscript. J.N.P.-L. performed all the statistical and spatial analyses with input from J.C.-B. All authors edited the manuscript.

Open Research

Data Availability Statement

All data used in this manuscript are publicly available. Main sources are provided in the WebPanel 1. R functions and examples for data analyses are publicly available at https://github.com/jesusNPL/FITBITs.

Acknowledgements

Support to this project was provided by the National Science Foundation (NSF) through the Macrosystems Biology & NEON-Enabled Science program (DEB 2017843). Further support was provided by the NSF BII ASCEND (DBI-2021898). The work is part of the FITBITs working group hosted by the National Center for Ecological Analysis and Synthesis.