Predicting ecosystem stability from community composition and biodiversity

de Mazancourt et al. (2013). Ecology Letters 16(5): 617-625. DOI:10.1111/ele.12088. Predicting ecosystem stability from community composition and biodiversity

Decomposing variation in community structure is… exactly as difficult as you would expect (taken from de Mazancourt et al.)

Will Pearse

Will Pearse

I’ve lost track of how many papers have tried to put forward a new way to understand ecosystem stability. I was drawn to this paper because it develops a novel conceptual framework that requires no more data than we already have, yet has greater explanatory power than other methods. The math is better, and so the model is better.

de Mazancourt et al. use data on individual species to predict what might seem like an abstract component of ecosystems – the coefficient of variation of community biomass. They’re not just predicting biomass or community composition, rather the stability of that composition over time. You’ve probably noticed I’m always desperate to link species’ ecology with how those species evolved; I wonder what the evolution of synchrony of environmental responses looks like. Do species that have coexisted for millions of years tend to be more synchronised? Or do they respond differently, and by responding differently ensure stable coexistence because they are occupying different niches (which reminds me of last week’s diversification limits paper)?

A fair bit of space is taken up with mechanisms by which observational error (which is an important component of their model) could have a biological interpretation. I’m not sure I quite follow, but I would be interested to know what effect intraspecific variation (which might be viewed as ‘error’) could have on all this. Intraspecific variation is a real (if, in my opinion, small) source of variation, and we might expect it to play a greater role in species that are more prevalent within a community (there are more of them, and so more opportunity for variation).

Finally, it is almost unbelievable that they were able to explain more than 70% of variation in the Texas dataset. So, seeing as how they’ve done unbelievably well in some datasets, and just plain-old-fashioned-very-well in others, why is there this variation? What is it about Texas that is so amenable to this method, and what makes Jena so different? I have literally no idea, and would be very grateful for ideas!

Lynsey McInnes

Lynsey McInnes

I told Will the paper I had chosen was too hard (what do you think – check it out here) and he came back with this one! Much harder! Although, so neat. I think I more or less get it. The authors set out to develop new theory as to why we ecosystems are so often more stable when they are more species rich. They neatly set out the conundrum of increasing richness stabilizing total community biomass, but at the same time destabilizing individual species abundances. Why does stabilization win out in most ecosystems?

The author list comprises a distinguished group of researchers working in this field and they bring together theory with four amazing time-series datasets. They are all set to make progress. And they do!  Basically (I believe) they find that stability is obtained through three distinct, but of course interacting, mechanisms: and I paraphrase, in more diverse communities you get a nicely complementary set of species that response asynchronously to environmental fluctuations so there is less chance of community collapse; in more diverse communities demographic stochasticity is weakened so you don’t get crashes of individual species; and finally more diverse communities, in effect, homogenize the intrinsic heterogeneity of an area through the provision of a set of species that altogether occupy all available ‘niches.’

The neat thing about this paper is that all of these mechanisms have been floating around in the literature already and have now been brought together into a single model and that model is verified with four independent empirical studies. The authors provide visually satisfying path diagrams to show how one gets from species richness to observed coefficient of variation of community biomass through each of the above mechanisms and they show the strength and direction of each effect.

Although this paper was a bit of dive back in time to one or two of my undergraduate lectures on overyielding and the insurance hypothesis, I did appreciate this paper and was thankful that it was well-written and well-explained. So many people work on elements of this puzzle, particularly motivated by curbing biodiversity loss into the future, but I think it’s grand projects like these – that set the situation up in a coherent framework – that might be most helpful in really demonstrating why and how diversity is beneficial to ecosystem stability. I imagine the authors’ heads are already teeming with next steps: climate change, evolutionary responses, invasive species, etc. One might want to know both how perturbing a system affects stability, but also if the relationship between diversity and stability stays the same during and post perturbation.


Biodiversity decreases disease through predictable changes in host community competence

Johnson, P. et al., 2013. Nature 494: 230-233. DOI:10.1038/nature11883. Biodiversity decreases disease through predictable changes in host community competence

Below, we give our first impressions of this article. Please comment below, or tweet Will or Lynsey (maybe use #pegejc). Think of this as a journal club discussion group!

Will Pearse

Will Pearse

Neither of us know anything about disease transmission, but we’d probably both claim to be ‘biodiversity scientists’ (whatever that means), so I picked this paper to shake things up a little bit. The paper is a very large-scale empirical demonstration that more diverse communities have fewer pathogens, implying that biodiversity in of itself helps stop disease transmission. It focuses on a single system, and they have measured one heck of a lot of frogs!

I have quite strong views about Ricklefs – I love his work, but I find his emphasis on pathogens strange. In Ricklefs’ view of the world, pathogens are the hidden puppeteers that control species diversity and many evolutionary dynamics. I argue that there is some truth to this, but it’s just too easy to say that observed patterns are because of something we can’t easily test. However, this paper demonstrates this effect rather clearly – not only is the rate of abnormalities reduced in more diverse systems when density is controlled for, but also the species found in the depauperate assemblages are the ones most tolerant of infection. I’m not convinced that pathogens are the only things driving this pattern; we might expect generally more ‘hardy’ individuals to survive in difficult environments, and if conditions are particularly harsh we might expect individuals to either die, or survive and thrive because they’re tolerant of the conditions but have reduced competition. However, this does paper does provide hope that we might actually be able to start measuring pathogen infection and explicitly linking infection with observed ecological patterns.

I’m not wildly enamoured with figure 3b – I think there’s a lot of scatter in the lower-left corner of the plot, and that the figure is influenced by a few outliers in the top-right. This means I’m not convinced there’s as strong a connection between snail density and amphibian density as the authors, but I’m willing to admit I’m being snarky. I’m much more excited about the experimental results that suggest this increase in transmission rate is not just a consequence of density changes – that there’s something about being in a diverse assemblage that gives a system more resilience to pathogens. This leads to a second question, I think – what’s the effect of increasing diversity in the pathogens? Is a more diverse pathogen community able to overpower a power diverse host community? While we’re on this subject (and perhaps this is something disease ecologists know a little more about), what is a pathogen community?

Lynsey McInnes

Lynsey McInnes

Wow, thanks Will! When I choose papers they fall right into his area of expertise – we he chooses papers, they fall right between either of our areas of expertise. Moving on…

I enjoyed reading this paper. It struck me that the authors took a hypothesis that has been being bandied about and went through a painstaking process to unpick it using a neat set-up incorporating field observations and lab and mesocosm experiments (I love mesocosms). In the process they got to play with a lot of diseased frogs.

My understanding is that support is accumulating for the idea that diversity in a community (here in terms of species numbers, presumably acting as a proxy for functional diversity) decreases disease transmission among individuals. Yet another plus point for maintaining diverse communities.

The authors stress that this effect is not just due to changing density of the species most able to transmit disease although the numbers point to this being part of the effect. Which is fair enough I guess. Rather, there is some kind of ‘dilution effect.’ How does the effect actually work? I don’t know enough about disease transmission to understand or really even speculate.

The authors state this ‘One possible explanation for the negative relationship between a host’s competence and its assembly order is that defences are costly and may incur trade-offs with resource investment in reproduction or dispersal. Indeed, studies in both eco-immunology and conservation support linkages between a species life history traits (for example, ‘pace of life’) and its vulnerability to infection or extinction, respectively.’ But I am not sure how this really relates to their findings and (I may have missed this), but I did think the analyses were concerned with host diversity versus assembly order (which is surely another dimension of the diversity issue?). My naïve thought process would lead me to think of the opposite: in diverse assemblages, hosts devote less to defense at the expense of reproduction, dispersal (indeed to competition with other species), that they are more vulnerable? I am without a doubt that someone could explain why my reasonings is wrong, but in the space of this paper, the authors did not manage to.

I was pleased to see the authors advocate considering more aspects of community competence than simply host diversity, (‘climate’, ‘resources’, ‘habitat’ – yay, ecology!), but did wonder why they did not consider any further biotic aspects (apart from snails P/A and density) of their communities that might affect their conclusions. What of interspecific competition? Are there any other unaccounted for disease hosts? I feel strongly like I’m missing something, but why do less resistant hosts dominate in low diversity communities? What are their traits?

And, the usual, from a macro-scale biologist. How do these results scale to different systems, different sites?

An interesting paper, timely, nuanced and thought-provoking.

A functional approach reveals community responses to disturbances

D. Mouillot et al., 2012. Trends in Ecology and Evolution 28(3): 167-177. DOI:10.1016/j.tree.2012.10.004. A functional approach reveals community responses to disturbances

Below, we give our first impressions of this article. Please comment below, or tweet Will or Lynsey (maybe use #pegejc). Think of this as a journal club discussion group!

Will Pearse

Will Pearse

I like the work of David Mouillot and Sébastien Villéger, and in my opinion they shouldn’t need to write a review paper explaining the suite of  approaches they’ve developed, because we should all be using them already. However, we’re not, and so they have, but we have the consolation prize of a really nice summary of some really nice approaches.

I’m what used to be called a community phylogeneticist – I use phylogenies to understand community ecological data. We are often (and rightly) accused of using phylogenetic data when trait data would probably be better for the question at hand – the phylogenetic middleman problem – and this paper describes the kinds of analyses we should do instead. However, we also come under fire because when we assume that phylogenetic overdispersion reflects competition, because it’s hard to map these observed patterns onto mechanisms (yes, I’m citing that paper again). The problem is that this exact critique can be applied to functional trait studies – if I find a community of species that are extremely dissimilar in terms of their functional traits, does that mean similar species have engaged in strong excluding competition, does it mean facilitation of dissimilar individuals is taking place, or does it mean something else? These issues are probably easier to tackle with trait data, but I still get worried when I start mapping trait patterns onto ecosystem processes.

Another thing that always confuses me in functional trait ecology is choice of traits. I don’t mean the boring (but important) “how many, and which, traits do we need to pick” issue, I mean what do we do when there’s variation in how important the traits we’ve collected are? These multidimensional trait spaces, to my mind, implicitly assume that all traits are equally informative. What if xylem diameter really doesn’t matter as much as specific leaf area? Can we weight the dimensions of these functional trait spaces to take this into account? How could we even detect that particular trait axes were more important? I think this is particularly important given we measure traits, in part, on the basis of what’s easy to collect – they’re what we hope are proxies for things we think are important in a system. Presumably someone has thought about this and I’ve just missed that literature – I’d be grateful for any links to papers.

There was very little about intraspecific variation in this paper, but then again it’s simple to incorporate into any of the measures: instead of abundances of each point in trait space, have lots of points (all of them close together, likely) for each measurement of a species. There was also little about choice of null models; this is something that does concern me slightly, because I’m unsure how to tell if we can compare measurements from a five-dimensional space to those from a ten-dimensional space safely. This probably reflects my dodgy maths! Moreover, in many other areas of ecology people get very concerned about the particular null model being used to test data – I just permute species identities for trait studies, and I can’t remember seeing anyone else doing anything else. Is that OK? Are there better ways of disentangling species composition and abundance?

Lynsey McInnes

Lynsey McInnes

Another strange pick from me that fell right into the community phylogeneticist niche that Will occupies. What a gift for him! Anyways, I was drawn to this paper because I like traits and functionality and am dubious (in a largely uninformed way) about most ways of defining communities. This paper kind of covers most of these topics…

I enjoyed this paper, but found it a bit difficult to follow. I think this is mostly due to me joining the discussion without having really thought that much about these ideas and issues beforehand. I liked the links to the niche/neutral debate, especially that the authors refrained from being really shitty about neutral theory and the contribution it can bring to thinking about ecological patterns and processes.

Some thoughts that came to mind…

Data availability/processing. The methods put forward in this paper look data/effort hungry. Which traits to use? How to score them objectively? Collating abundance data? Worrying about missing rare species/traits (although maybe this doesn’t matter)? Worrying about trait x trait interactions? Minefield! But in terms of providing early warning signals for community functioning probably worthwhile?

My next worry is how common is it to have this detailed information on shifting abundances following disturbance in order to benefit from these early warning signals of impending local extinction? Perhaps it is fairly common (or at least feasible) in communities where we try hard to obtain this information. Also (and there probably exists a whole research field on this), maybe we want to be focussed on some emergent properties of communities that can be quantified without monitoring abundances of all species involved, i.e., if functional traits are what is important rather than species diversity or identity then don’t we want some way of measuring community functionality that is not dependent on being able to monitor species’ abundances…

All this emphasis on traits over species diversity made me wonder how these techniques and approaches fit within the whole debate on species identification via traits vs. DNA barcodes? If a species is largely functionally redundant, do we care about its species’ status? I’m being artificially simplistic, of course, and I fully believe there is a place for these functional approaches and DNA barcoding approaches depending on both the aim of the exercise (discovering diversity vs. predicting responses to disturbance being an obvious distinction) and the nature/location of the community.

I’ve always been on the edge of the field of community phylogenetics but never had the guts to dive right in. But it does seem that a big criticism of CP (beyond the maze of dodgy metrics) is the focus on species as the be all and end all of most measures, rather than figuring out how trait differences (big and small) influence the different community structures that we observe (filtered, overdispersed, etc.). CP would probably benefit from incorporating some of this focus on traits, no?

The focus here is clearly on quite rapid timescales but with the increasing recognition that ecological and evolutionary timescales have a big chunk of overlap, I found it interesting that the authors didn’t devote much time to speculating on any possible evolutionary responses of species that decline in abundance. Yep, the scope is probably minimal and the trait values they would ‘need’ in the truncated functional space are likely already ‘taken’, but even so…

And finally, back to what is a community? It would be fun if these kinds of methods could somehow be adapted to identify clusters of function that somehow correspond meaningfully to the abstract idea of a community that most of us already hold in our heads…

%d bloggers like this: