Should we conserve phylogenetic diversity?

Analysis from data on ecologically-relevant traits from >15,000 vertebrate species show that maximizing phylogenetic diversity results in an average gain of 18% of functional diversity relative to random choice, but this gain is not significant

Go to the profile of Florent Mazel
Jun 09, 2018
Upvote 4 Comment

The full paper is here

Phylogenetic diversity and conservation biology

There is clearly academic interest in phylogenetic diversity (PD) as a metric of conservation prioritization: Dan Faith’s seminal 1992 paper has been cited more than 2329 times as of June 2018 (data: Google Scholar). Adoption by conservationists, however, has been delayed in the extreme, in part due to the lack of strong conceptual and empirical links between phylogenetic diversity and more traditionally valued aspects of biodiversity. Beginning in 2016, Arne O. Mooers and Caroline Tucker organized a diverse set of young scientists to focus on the links between PD and values relevant to conservation. With support from the Synthesis Centre of the German Centre for Integrative Biodiversity Research and the Canadian Institute for Ecology and Evolution, the working group met three times over two years. I started my first post-doc with the objective to lead some of this necessary synthesis work.

Why should we conserve phylogenetic diversity? 

During our first meeting in Ottawa (Fig. 1), the working group identified that a key (but often unstated) argument for conserving PD is that by protecting more PD, we should also protect a greater amount of total trait (or functional) diversity (FD). Protecting greater trait diversity in turn might lead to the maintenance of more biological goods and ecosystem services for direct use, increased ‘option value’ (i.e. biological goods useful in the future), and the raw material for future biodiversity production via evolution. This central hypothesis, which we named the ‘phylogenetic gambit,’ has been widely embraced by some academic biologists but had, surprisingly, never been tested. The aim of my post-doc became to test just that. The question was simple: if we chose a subset of species to maximizing the phylogenetic diversity contained in that group, do we also capture more functional diversity (FD) compared to if we just selected a random subset of the same number of species?

Figure 1. The working group during the first meeting in Ottawa
Figure 1. The working group during the first meeting in Ottawa

Troubling theoretical results…

I started exploring this question using simulations. We had strong expectations that the PD-maximization strategy should be better at capturing FD than random choice. But we were wrong, and that was pretty exciting: in some plausible evolutionary scenarios, maximizing PD could actually lead to lower FD than a random choice. This somewhat troubling finding had the potential to invalidate the PD-based conservation program (Mazel et al. 2017).

Empirical test

I took these simulations results to the additional meetings were held in Leipzig (Figure 2). The real question in all of our minds was, 'what does the phylogenetic gambit look like in the real world?' We identified global datasets of the distribution, traits and phylogenies for >15,000 mammals, birds and fish species. When I applied the same question (does selecting species to maximize PD better capture PD compared to a random set?) to this data, I found that prioritizing the most phylogenetically diverse set of species did lead to an average gain of 18% more trait diversity compared to applying the same conservation effort without considering phylogeny (Figure 4). The phylogenetic gambit seemed to pay off in the real world.  However, this average gain masked an important caveat: the PD prioritization scheme on our data was quite unreliable - in fully 36% of all our runs, it captured less FD than a scheme that ignores phylogeny.  The troubling feeling remained.

Figure 2. The working group during the second meeting in Leipzig
Figure 2. The working group during the second meeting in Leipzig

What’s next?

So, what’s next? Shall we abandon or pursue the PD-based conservation program? Our empirical test suggests that, while global conservation initiatives focusing on PD will, on average, capture more FD than a random strategy, it might be considered fairly risky. However, we are not ready to say more: our test needs to be repeated in other taxa and at other spatial scale, and, critically, with other and more trait data.  Given how important plants are to ecosystems, they should be the next target.  There seems no other way to narrow the uncertainties of our results.



Faith, D. P. (1992). Conservation evaluation and phylogenetic diversity. Biological conservation61(1), 1-10.

Mazel, F. et al. 2017. Conserving Phylogenetic Diversity Can Be a Poor Strategy for Conserving Functional Diversity. - Syst. Biol. 66: 1019–1027.

Go to the profile of Florent Mazel

Florent Mazel

Post-Doc, University of British Columbia


Go to the profile of Daniel P Faith
Daniel P Faith 24 days ago

Mazel provides a welcome “behind the paper” piece. How they address its title “Should we conserve phylogenetic diversity?” reveals some key weaknesses of their idiv project. I’ll contrast my answer to that question (1) with their very different one (2): 

(1) Should we conserve phylogenetic diversity? The answer is “yes”. Faith (1992) from the outset stressed that PD should not be expected to recover functional traits prone to convergent evolution. However, PD has a well-corroborated link to general feature diversity, based on many tests over the years. The link to feature diversity is the basis for addressing the value PD was designed to address – and perhaps the oldest stated value of biotic diversity – “option value” (for discussion and links to corresponding papers, see ). This has promoted applications (e.g. providing a rationale for EDGE and for current biodiversity policy work in NSW).  IPBES (noting that well-corroborated relationship) recently has used indicators of expected PD loss, over 6 major taxonomic groups, to assess “option value” provided by biodiversity (Davies et al in press).

(2) Should we conserve phylogenetic diversity? Their answer seems to be “maybe not”.  They claim little take-up of PD by conservationists “in part due to the lack of strong conceptual and empirical links between phylogenetic diversity and more traditionally valued aspects of biodiversity.” To fill this supposed gap, their idiv group focuses on functional traits for “the links between PD and values relevant to conservation” Further, they now claim the “key (but often unstated) argument for conserving PD is that by protecting more PD, we should also protect a greater amount of total trait (or functional) diversity (FD).” Their “ecologically relevant traits” for an FD convex hull are just 4 traits for birds and mammals, all probably prone to convergence. Arguably, PD should not reflect such FD (see also Faith in press).

Conclusion: The idiv group appears to have ignored the well-corroborated link of PD to feature diversity, and the long history of option value with links to the biodiversity crisis (see discussion/papers in ). Instead, in their paper’s first paragraph, they interpret the biodiversity crisis only in terms of ecosystem functions/services or moral values. As I have argued elsewhere (Faith 2018), appreciation of biodiversity option value is critical to addressing the biodiversity crisis. PD provides an effective way to reflect these values.


Davies, K., A. Rajvanshi, Y. Yeo-Chang, A. Gautam, A. Choi, A. Masoodi, C. Togtokh, H. Sandhu, H. Husain, J. Cho, et al. In press. Chapter 2 in M. Karki and S. Sonali, editors. Nature’s contributions to people and quality of life. IPBES, 2018: Regional and subregional assessment of biodiversity and ecosystem services for Asia and the Pacific. Secretariat of the Intergovernmental Platform for Biodiversity and Ecosystem Services, Bonn, Germany.

Faith DP (1992) Conservation evaluation and phylogenetic diversity. Biol Conserv 61:1–10.

Faith DP (2018) Biodiversity’s option value: A comment on Maier (2018). Ambio.

Faith DP (in press) Phylogenetic diversity and conservation evaluation: perspectives on multiple values, indices, and scales of application. In: (R. Scherson and D.P. Faith eds.) Phylogenetic Diversity. Applications and Challenges in Biodiversity Science. Springer. 

Go to the profile of Daniel P Faith
Daniel P Faith 18 days ago

Additional comments : Unreliable perspectives on phylogenetic diversity

Faith (1992) proposed a fundamental link between phylogenetic diversity (PD) and general “feature diversity”. PD sees shared features (among species) as explained by shared ancestry. That PD model is well-corroborated, given the many tests over the years over multiple sets of features and multiple taxonomic groups. Thus, the PD-features link has been well-tested. This link in turn connected PD to the long-recognised fundamental benefit/value of biodiversity to humans: “option value” (for overview and links to papers; see This well-corroborated relationship is the basis for use of PD by the Intergovernmental Platform for Biodiversity and Ecosystem Services (IPBES) as one indicator for “maintenance of options” (one of “nature’s contributions to people”). Faith (1992) also warned that functional traits prone to convergent evolution would not be captured reliably by PD (in related early work, Faith proposed an alternative framework for functional trait diversity.

Mazel et al’s title reflects their key claim that PD “captures functional diversity unreliably”. This might appear to be simply a re-discovery (sadly, without referencing) of what Faith had already documented 25 years ago. But the situation is even more complicated than that. Mazel et al. begin by indicating that “FD” (“functional diversity”) happens to be their chosen term for the vast diversity of “form and function” (so appearing to be compatible with Faith’s broad “feature diversity”).  This would give the impression that they have shown that PD “captures feature diversity unreliably”. However, Mazel et al’s actual analysis defines FD as a few nominated “ecologically relevant” traits. FD for mammals and birds is only 4 ecological traits (diet, body mass, activity cycle, and foraging height).  These are perhaps prone to convergent evolution – exactly the sort of ecological traits Faith saw as weakly linked to PD. Through this FD word-play, Mazel et al. give the impression that Faith’s argument that PD should capture feature diversity is equivalent to a requirement that PD should capture this FD. 

Mazel et al. also want the reader to believe that “it is widely argued” that, “to keep humanity’s options open”, we should seek to maximize the “ecologically relevant” traits making up “FD”.  In reality, this does not correspond to the actual widely-argued idea that conservation of overall feature diversity, through PD, keeps humanity’s options open. This distinction makes it clear that the case for PD conservation should not be equated with Mazel et al’s “phylogenetic gambit” based on FD. 

Mazel et al. carry out a test to see if PD recovers this FD. PD does pretty well, but sometimes performs worse than a random set of species.  Mazel et al. conclude that PD is unreliable. Again, pursuing such a test might seem odd, given Faith’s past warning about such ecological traits, and given the corroboration of PD already achieved for lots of feature data. But the result of their analysis is odder still - it turns out that it is the test itself that is unreliable. 

To see this, let’s look at the basic mathematics of PD. Suppose that PD has perfect recovery of features – all shared features are explained by shared ancestry. Maximum PD sets then are also maximum features sets. Any “test” should show good PD performance. But the Mazel et al test perversely would say the opposite.  The reason is that the test looks at the recovery of just (say) 4 nominated features/traits, while PD is trying to capture (say) 4 million features.

A little schematic drawing gives a feel for the problem (see figures 1 and 2 at ). In Fig. 1, we have a tree with 12 species. A single “FD” trait has two states, 1 and 2.  We can directly calculate results for random, max PD, and max FD sets. Fig 2 (comparable to Mazel et al’s Fig 1) shows that max PD sets, for set-size up to 5 species, capture lots of feature diversity, but score “0” in the recovery of this one trait (the max PD set in each case does not include any of the branches leading to a “2”). Random sets have a much greater chance of capturing this particular trait diversity. The result is that the Mazel et al test would report PD as much worse than random (Fig.2) – even when PD sets in fact have captured (maximised) feature diversity perfectly! I conclude that it is Mazel et al’s test, not PD, that is “unreliable”. This test should not be used. Mazel et al’s conclusions should be disregarded.

What lessons can we learn from this misadventure? I think one lesson is that a more scholarly approach to past work is needed. Their paper is one from their sCAP working group, funded by sDIV. A cornerstone of that sDIV work seems to be a number of critical miss-representations about PD history. For example, they sometimes give the impression: that PD was defined at the community-level; that the PD rationale is about functional traits; and that PD has not been tested (for discussion, see At the same time, they have ignored some core PD history, including the shared-ancestry PD model, the PD “calculus”, and the problem of convergent functional traits (for other related problems with their history see also ).

Mazel et al refer to the “phylogenetic gambit” as requiring that “maximizing PD yields more FD than a random strategy”. Given that this notion is ill-conceived, perhaps we now can find a new use for their evocative phrase. I suggest that the real “phylogenetic gambit” at play here might be the consistent miss-representation of PD history, which then favours other contrary arguments. Let’s hope that this is a phylogenetic gambit that fails.