Ask us (well, Brian) anything: generalities about trait-based community ecology?

Every year we invite you to ask us anything! Today’s question is from Eric Charnov, and is for Brian (paraphrased, click that last link for the original):

What are some of the general rules for community ecology that have emerged from the functional traits research program?

Brian’s answer:

So in my 2015 post on “Steering the trait bandwagon” I mostly identified gaps and shortcomings in where traits have gone since my 2006 paper you refer to (and of course several other key papers and authors advocating use of traits). I still think that every time someone publishes a paper just saying “we measured some traits on a group of species” people are at risking of thinking the trait-based approach is intellectually shallow. I also stand by most of my “steers” in that post (e.g. to get more multivariate, to get traits by environment, to work more on the individual level). But I read this question as an invitation to highlight advances or accomplishments in traits (or at least I’m going to choose to read it that way) and provide the positive side of the assessment. So some key accomplishments in the trait world:

  1. We are building large databases of traits, most especially plants (of which the largest are into millions of records, tens of thousands of species, and dozens of traits including BIEN and TRY but there are all kinds of regional and more specialized databases). But also  macroinvertebrates, corals, fungi, lizards, butterflies, deep sea vent fauna, fish (here and here), ants (here and here), and etc. Building a database is of course a means, not an end, scientifically. But I have been impressed at the rapid explosion and quality of trait databases.They are a key step towards generality and a positive reflection on the movement to open data.
  2. Assembly can be better described by traits than species – Although far from the only example of this claim, I think the paper led by Julie Messier, my former graduate student, now faculty at Waterloo, is still the most powerful example of this. Look at the figure below from Messier et al 2010. The graphs represent kernel densities of a trait distribution. Left to right gives 3 sites along a gradient in Panama (North/Atlantic/wet to South/Pacific/dry). At each site, for 4-8 plots (30m x 30m) every individual was sampled for a number of traits including leaf-mass-per-area (LMA). The thin lines represent different plots (~100 m apart) at the same site. The thick line represents the average. While there is some plot-to-plot variability, it seems clear that not just the mean but the variance and skew of the trait distribution is conserved across plots within a site presumably experiencing identical climatic conditions (but varies across sites with varying climatic conditions). I find this result especially impressive as it is in the tropics and the overlap of species between plots within site is only about 13-33%. In other words, 67-87% of the species are different between plots, and yet the trait distribution is strongly conserved. If we were to take a classical species-based approach we would have to throw up our hands and say species composition is primarily just stochastic outcomes of dispersal, yet with a trait based approach we see some real ecophysiogical determinism coming in that is missed with a species approach.Example of trait filtering
  3. Abundance and demographic rates like mortality can be better predicted by traits than species – Closely related to #2, if traits are being “filtered” then traits must be tied to demographic processes and outcomes. Traits have been successful at predicting relative abundance of different species (Shipley 2006, Laughlin 2012), and species demographic rates (Poorter 2008, but see Yang 2018 for reasons why predicting rates for a single species out of context has real limits as also suggested in my steering traits post). Somewhat more promising as a future direction, traits are good at predicting outcomes of competition such as predicting impacts of neighbors on growth and mortality of a focal individual better than taxonomic identity or relatedness (Uriarte 2012, Fortunel 2016) and the same for coexistence mechanisms (Kraft 2015). There are many more I could have cited. Contrast that success to the success of traditional species-interaction/community-matrix approaches to predicting community dynamics outcomes such as relative abundance or effects of competition (basically close to zero). The community matrx is parameter hungry (~(n^2)/2 paramaters for n species). And the only way to parameterize is is to phenomenologically measure the nature of each pairwise interaction. Not really going to to happen. The fact that we can measure traits of individual species (in isolation – so ~ n parameters not n^2) and use them to predict the outcomes of species interactions is exciting and relatively unprecedented.
  4. Functional diversity is interesting and clearly separable from taxonomic diversity and relevant to ecosystem function – it has been commonplace for at least a decade now to note that we should supplement taxonomic diversity (be it richness or a measure that incorporates evenness) with phylogenetic and functional diversity. I’m not going to address phylogenetic diversity here, but I would claim that functional diversity (especially taking residuals after controlling for taxonomic diversity) has provided some really interesting and unexpected results. Lamanna 2014 showed that taxonomic and functional diversity follow taxonomic diversity in being higher in the tropics, but at a gamma or regional scale functional diversity is actually higher in temperate zones. Petchey 2007 found that functional redundancy in individual bird assemblages is lower than expected by chance, again suggesting environmental filtering might be acting but then when looked at over 20 years, functional redundancy is being lost just as fast as taxonomic richness so there is no evidence of functional redundancy buffering ecosystem functioning against species loss. Tilman 1997 showed that functional diversity and functional composition are both better than taxonomic diversity in predicting ecosystem function. And Dynamic Global Vegetation models (a key improvement over GCMs in predicting climate change) are increasingly moving to using traits (van Bodgeom 2014). Again this is just the tip of the iceberg, but it suggests that taking a functional diversity view of the world adds new information beyond species richness. Functional composition and diversity are also key to model ecosystem functioning.
  5. There are some recurring syndromes or axes of trait variation – understanding the diversity of 10,000,000 species is a core challenge. Finding major, repeatable axes of variation is a key tool for reducing this dimensionality. The finding of the role of body size and a fast-slow-continuum in life history has been a major advance for example. A recent (Diaz 2016) paper confirmed that in plants a whole plant and plant organ size dimension captures the most variation with much of the rest captured by a fast-slow-leaf dimension (also see Westoby 1998, Wright 2004). I am sure there are some more axes out there related to, e.g. reproduction and seeds or wood and height as well as in other organisms but I haven’t seen those wholly and decisively nailed down yet to the same degree as the first two axes I mentioned in plants.

That would be my summary of some of the key findings and results in the trait world (most but not all since 2006). For sure we have a lot of work still to do. But at a minimum, I think traits have established themselves as an approach that brings something new, different and valuable that hasn’t been there in community ecology before.

What do you think? Have traits changed or improved our understanding of ecology? Are there key advances (or papers) I left out?


13 thoughts on “Ask us (well, Brian) anything: generalities about trait-based community ecology?

  1. Hi Brian; many thanks for the broad sketch and a beginning reading list for communities and ecosystems; its mostly new to me [ which is why I asked] and sounds quite promising. much winter reading ahead.
    As for the’ slow/fast continuum’ in life histories I am less excited about recent work.
    The slow/fast terms originally replaced ‘r/k’ selection as a descriptor and basically referred to an adult body size dimension; then folks showed for mammals that there were correlations among the demographic residuals for various fitted allometries for life history variables [ all this in the 1980s]. Early work was descriptive, with some small side glances to life history theory. [ Maynard Smith was an external examiner on one early doctoral thesis, and told me afterwards….’ but there aint no theory here’]. I found the data very stimulating, but also puzzling; I thought long and hard about it.
    My 1991 mammal life history model [ here: used life history theory and demographic stability in the face of power function [ metabolic] production/growth to predict both the allometries and the correlations among residuals; it worked well for mammals; (but not perfectly of course.)

    recently there has been an explosion of descriptive papers , with many taxa [ of course, there are many more data sets and statistical programs]. Some of the patterns found in mammal ‘slow/fast’ repeat in other taxa, but the main theme seems to be finding how other taxa are different; and its now done with more fancy statistics. and more dimensions: the 2 main themes are that adult demography and adult body size are one slow/fast dimension and everything about offspring number per unit time is another.People seem to have forgotten that Stearns showed the exact same pattern in his 1977 ARES paper criticizing r/k selection[ eg, fish don’t generally show a correlation across species for adult size and egg size, while mammal offspring are close to proportional to their mom’s size; so fish clutch [sic] size increases with body size and the reverse for mammals].

    But mostly we have returned to the old descriptive days( with a vengance}, with little if any link to life history theory.
    this troubles me since formal life history theory played such a big role 30 yrs ago! Without theory we will never understand the similarities and differences in various data sets, regardless of their size.

    • Thanks for this inside, expert view on life history trade-offs. I’m about to start writing a chapter in my macroecology book on life history so this is all excellent information!

      I think much of what you said could be applied to the quest for trait trade-offs (for some reason they call them economic dimensions in traits). It is primarily empirical with little theory and yes body size and then fast/slow and yes reproduction can show up as a third dimension. The interesting/complication issue is since plants have modularity one could imagine fast/slow roots, fast/slow wood, fast/slow branching/growth, fast/slow leaves. How do these relate to each other? fast on one=fast on all? or not? Peter Reich ( says they are all coordinated or parallel. Although in truth to my eye we have not yet got decisive, general views of fast/slow outside of leaves.

  2. Super interesting Brian, thanks for the summary, I learned a lot about recent developments in this area from your post. Floral evolution and interactions with pollinators has rather been neglected within this area of trait-based community ecology, or perhaps has stood apart from it, with some notable exceptions, particularly in the network literature. I’d like to see more of that integrated into the kind of work you are describing. One of the reasons why it’s not been well integrated to date is that it’s not always clear what traits to measure: seed mass or leaf area are relatively straightforward (with the emphasis on “relatively”), but how do you “measure” a flower? No single trait correlates with a given type of pollinator.

    So I was struck by your section title “There are some recurring syndromes or axes of trait variation”. Pollination syndromes have, historically, been one of the favoured ways of understanding how flowers evolve to particular groups of pollinators. But recent work by myself and others has suggested that on average only about a third of the flowering plants can be slotted into traditional syndromes, with floral phenotypes of most species sitting outside of the multivariate phenotypic boundaries of “expected” syndromes. That’s partly because a lot of species are highly generalized in their interactions with pollinators, partly because novel types of interactions keep being discovered (I estimate that we have data on the pollinators of no more than about 10% of the flowering plants) and partly because the pattern-recognition parts of the human brain insists on looking for examples that fit preconceived ideas, and studying them, rather than taking an unbiased sample of the plants in whole communities.

    No real question here, but I wondered if you had given it any thought?

    • I am obviously a trait-biased person, but I do think traits could be a constructive approach to pollination. To me the core concept is to get beyond what my colalborator and co-author Evan Weiher called nomenclatural ecology which ultimately is just long lists of species and hard to generalize or predict or extrapolate. Which as you noted is badly needed. We have to conserve species that we know nothing nomenclatural about their pollinators. Or to quote from our 2006 paper: “the trait-based statement ‘compact plants with canopy area !30 cm2 and small or absent leaves are restricted to marshes with ~18 mg gK1 soil P’ is more useful than the nomenclatural statement ‘Campanula aparinoides is found only in infertile habitats.’ Statements about traits give generality and predictability, whereas nomenclatural ecology tends towards highly contingent rules and special cases.”

      The functional trait world also went through discrete syndromes too. (EG evergreen needle vs evergreen broadleaf vs water deciduous broadleaf vs temperature deciduous broadleaf). But as you so rightly note the real world is continuous and high dimensional. Which makes it likely many species fall outside of on the boundaries of our boxes. Those syndromes say as much about human cognitive perception as they do about the biology. So it is important to move from discrete to continuous.

      So in my mind traits are solutions to the problems of discreteness (arbitrary boxes or groups) and nomenclatural (knowledge is organized by lists of species we’ve studied which is inherently hard to generalize).

      With relation to pollination my thought centers on the notion of interaction traits. Partly because of its roots in ecosystem function and partly because they are standardized measurements, the plant trait world has been very focused on physiological traits. And as some of my citations in the blog show physiological traits (and size traits) can be fairly predictive about outcomes of competition in plants. But this is still an extremely limited world view. Reproduction traits (shades of Grime there) and interaction traits (shades of MacArthur there) are every bit as important as physiological traits.

      But very broadly every type of species interaction should have a set of traits that are relevant to that interaction if we think out of the physiological trait box. I have been on the committees of several students who looked at pollinator traits (petal color, % area of petal with UV markings, petal length & width, sepal morphology, nectar sugar concentrations, etc). And equally there should be traits on the pollinator side (beak length, wing buzz frequencies, wing ratios that get at ability to hover or not, % diet from nectar, pollen and other sources, pollen carrying morphology (hairs on head, pollen baskets,etc) and more. It ought to be possible to get these to come together towards a general and predictive rather than nomenclatural view of pollination. I’m not aware of any papers that got published on the plant pollinator traits from those committees I was on. and I certainly don’t know of any paper that has pulled the animal and plant stories together in a trait-centric way.

      Of course most of what I said are not new ideas in pollination ecology. I remember reading a paper in grad school by Armbruster (so 1990s) on pollination syndromes that started with rigorous trait measurements. But the end goal was discrete pollination syndromes (if I recall correctly cluster analysis entered somewhere). I do think the idea of continuous interaction traits in a multivariate context could provide a constructive lens for framing this work though.

      • Thanks for the detailed response Brian, that’s given me a lot to think about. But one of the hurdles to predictability of flowers, at least as far as continuous traits go, is the phylogenetic and life history diversity of pollinators. I recently calculated that as many as one in 10 described terrestrial animals might act as pollinators – that’s a lot of diversity! Even within one group, the bees, we see a range of motivations for visiting flowers: it could be for nectar, pollen, shelter, warmth, scent, resin, sexual deception, or food deception. So the trait “rewards offered by flowers to bees” is not at all continuous or quantifiable along a single axis.

        I need to ponder some more on this.

      • All true – I would only add the intention (whether it succeeds or not) of traits is to deal with that enormous diversity by moving out of the species by species mindset. And my biggest gripe about how traits have been implemented to date is as a bunch of independent univariate measures – I absolutely think traits have to be treated in a multivariate fashion and you give an excellent example of why.

      • A little late to the party, but…

        Following from, “The functional trait world also went through discrete syndromes too. (EG evergreen needle vs evergreen broadleaf vs water deciduous broadleaf vs temperature deciduous broadleaf)….”

        Isn’t that still where traits show up in major GCMs like CLM/ED/FATES? Seems like those models are still based on PFT = Leaf habit or leaf lifespan, possibly also with SLA or Nitrogen content. But leaf habit, from my (limited) experience isn’t really a great predictor of water use efficiency, for example, (or photosynthesis), because the traits that maintain water use don’t neatly track leaf habit.

        So I guess my impression is that the modeling world has fixated on leaf habit (it’s easy) and in doing so misses the promise of PFTs. Or perhaps I’m being the times on current model states?

      • skipvb – you’re right the major GCMs are still at functional groups.

        But some of the DGVMs (more sophisticated global vegetation focused models) are increasingly switching to continuous traits (see Van Bodegom reference in the main post) and those often serve as precursors to what goes into GCMs (not so much the weather forecasting ones that don’t have vegetation as more than albedo as I understand it but, the climate time scale model versions of GCMs). But I probably could have been more clear on the DGVM/GCM distinction.

  3. Late follow-up on this, but I thought it’s worth pointing out: fisheries ecology and marine community ecology has been going through a trait-based modelling renaissance with the rise of size-spectrum approaches for modelling the dynamics of whole communities. These models are based on the idea that most marine species have very predictable growth patterns, and that the trophic level of a given individual is much better predicted by the size of the that individual, compared to its species identity, and can be used to predict the dynamics of whole communities. It’s interesting because it’s being driven heavily by theorists. It seems to work much better than the old “ecopath” approach for modelling marine communities, which required detailed physiological data on huge number of species plus a lot of tweaking of numbers to get sensible community dynamics.

    Traits make their way into the size-spectrum models through the maximum size that each species can grow to, habitat preferences, and between-species variation in growth rates versus ability to avoid danger (a risk-reward trade-off). Julia Blanchard, Ken Andersen, and Axel Rossberg have been some of the theorists leading the drive to developing this approach, and Ken Andersen just published an Monograph in Population Biology called “Fish Ecology, Evolution, and Exploitation: A New Theoretical Synthesis” that’s a great jumping-off point to it.

    • Thanks so much for the info/references.

      For sure if you can only know one trait (beyond that it is a fish instead of a tree) body size has to be number one. Great to hear about how they are making such rich use of body size.

    • Hi Eric; thanks for pointing out the marine ecology work on size structured communities. It is indeed true that theorists have played important roles[ and Ken Andersen’s recent book looks very interesting]….Fisheries has a long tradition of quantitative models for pops, etc, and body size has long been recognized as a leading factor. Fisheries folks have also made much productive use of life history evolution modelling, and [ as you note] integrating it into community level patterns. Quantitative rules about body size , growth parameters, and mortality rates have been known and used in fisheries since the late 1950s [ reviewed in my 1993 book].
      This is true for freshwater fish also, and the work of Earl Werner, James Gilliam, and Gary Mittlebach is particularly important, going back 40 yrs; see
      and elsewhere. See particularly the 1984 ARES article for an overview;
      And in life history terms the themes of growth vs risk-of-death are the keys.

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.