#ESA100: The next dimension in functional ecology

The third day of ESA talks saw an interesting session on functional ecology (Functional Traits in Ecological Research: What Have We Learned and Where Are We Going?), organized by Matt Aiello-Lammens and John Silander Jr.

As outlined by McGill and colleagues (2006), a functional trait-based approach can help us move past idiosyncrasies of species to understand more general patterns of species interactions and environmental tolerances. Despite our common conceptual framework that traits influence fitness in a given environment, many functional ecology studies have been challenged to explain much variation in measured functional traits using underlying environmental gradients. We might attribute this to a) measuring the ‘wrong’ traits or gradients, b) several trait values or syndromes being equally advantageous in a given environment, or c) limitations in our statistical approaches. Several talks in this organized session built up a nuanced story of functional trait diversity in the Cape Floristic Region (CFR) of South Africa. Communities are characterized by high species but low functional turnover (Matt Aiello-Lammens; Jasper Slingsby), and only in some genera do we see strong relationships between trait values and environments (Matt Aiello-Lammens; Nora Mitchell). Nora Mitchell presented a novel Bayesian approach combining trait and environmental information that allowed her to detect trait-environment relationships in about half of the lineages she investigated. These types of approaches that allow us to incorporate phylogenetic relationships and uncertainty may be a useful next step in our quest to understand how environmental conditions may drive trait patterns.

Another ongoing challenge in functional ecology is the mapping of function to traits. This is complicated by the fact that a trait may influence fitness in one environment but not others, and by our common use of ‘soft’ traits, which are more easily measurable correlates of the trait we really think is important. Focusing on a single important drought response trait axis in the same CFR system described above, Kerri Mocko demonstrated that clades of Pelargonium exhibited two contrasting stomatal behaviours under dry conditions: the tendency to favor water balance over carbon dioxide intake (isohydry) and the reverse (anisohydry). More to my point, she was able to link a more commonly measured functional trait (stomatal density) to this drought response behavior.

Turning from the macroevolutionary to the community scale, Ben Weinstein evaluated the classic assumption of trait-matching between consumer (hummingbird beak length) and resource (floral corolla length), exploring how resource availability might shape this relationship. Robert Muscarella then took a community approach to understanding species distributions, testing the idea that we are most likely to find species where their traits match the community average (community weighted mean). He used three traits of woody species to do so, and perhaps what I found most interesting about this approach was his comparison of these traits – if a species is unlike the community average along one trait dimension, are they also dissimilar along the other trait dimensions?

Thinking of trait dimensions, it was fascinating to see several researchers independently touch on this topic. For my talk, I subsampled different numbers and types of traits from a monkeyflower trait dataset to suggest that considering more traits may be our best sampling approach, if we want to understand community processes in complex, multi-faceted environments. Taking trait dimensionality to the extreme, perhaps gene expression patterns can be used to shed light on several important pathways, potentially helping us understand how plants interact with their environments across space and time (Andrew Latimer).

To me, this session highlighted several interesting advances in functional ecology research, and ended with an important ‘big picture’. In the face of another mass extinction, how is biodiversity loss impacting functional diversity (Matthew Davis)?

McGill, B. J., Enquist, B. J., Weiher, E., & Westoby, M. (2006). Rebuilding community ecology from functional traits. Trends in ecology & evolution, 21(4), 178-185.

A good null model is hard to find

de Bello, F. The quest for trait convergence and divergence in community assembly:are null-models the magic wand? Global Ecology and Biogeography, (Global Ecol.Biogeogr.) (2012) 21, 312–317

Ecologists have always found the question of how communities assemble to be of great interest. However, studies of community assembly are often thwarted by the large temporal and spatial scales over which processes occur, making experimental tests of assembly theory difficult. As a result, researchers are often forced to rely on observational data and make inferences about the mechanisms at play from patterns alone. While historical assembly research focused on inferring evidence of competition or environmental filtering from patterns of species co-occurrence, more recent research often looks at patterns of phylogenetic or trait similarity in a community to answer these questions. 

Not surprisingly, when methods rely heavily on observational data they are open to criticism: one of the most important outcomes of early community assembly literature was the recognition that patterns that appeared to support a hypothesis about competition or environmental filtering could in fact result by random chance. This ultimately lead to the widespread incorporation of null models, which are meant to simulate patterns that might be observed by random chance (or other processes not under study), against which the observed data can be compared. Patterns of functional and phylogenetic information in communities can also be compared against null expectations to ensure that patterns of phylogenetic or functional over- or under-dispersion can't arise due to chance alone. However, while null models are an important tool in assembly research, they are sometimes as the foolproof solution to all of its problems.

In a new paper by Francesco de Bello, the author states frankly “whilst reading null-model methods applied in the literature (indeed including my work), one may have the impression of reading a book of magic spells”. While null models are increasingly sophisticated, allowing researchers to determine which processes to control for and which to leave out, de Bello suggests that the decision to include or omit particular factors from a null model can be unclear, making it difficult to interpret results or compare results across studies. Further, results from null models may not mean what researchers expect them to mean.

Using the example of functional diversity (FD; variation in trait values among species in a community), de Bellow illustrates how null models may have different meanings than expected. Ideally, a null model for FD should produce random values of FD, against which the observed values of FD can be compared. Interpreting the difference between the observed and random results can be done using the standardized effect size (SES, the standardized difference between the observed and randomly generated FD values); SES values >0 show that traits are more divergent than expected by chance, suggesting competition structures communities. If SES<0, traits are more convergent than expected by chance, suggesting environmental conditions structure communities. Finally, if SES ~0, then trait values aren’t different from random. However, de Bello shows that the SES is driven by the observed FD values, because the ‘random’ FD values are dependent on the pool of observations sampled. This means that the values the null model produces are ultimately dependent on those observed values, despite the fact you plan to make inferences by comparing the null and observed values as though they are independent. For example, consider the situation where you are building a null model of community structure for plant communities found along two vegetation belts. If the null model is constructed using all the plant communities, regardless of the habitat they are found in, the resulting null FD value will be higher, since species that are dissimilar and found in different vegetation belts are being randomly selected as occurring in a community. If null models are constructed separately for both vegetation belts, the null FD value is lower, since species are more similar. The magnitude of the difference between the null model and the observed values, and further, the biological conclusions one would take from this study, would therefore depend on which null model was constructed.

from de Bello 2012, illustrating how combining species pools (right) can lead to entirely different decisions about whether communities are convergent or divergent in terms of traits than when they are considered separately (left, centre).

De Bello’s findings make important points about the limitations of null models, particularly for functional diversity, but likely for other types of response variable. The type of null model he explores is relatively simplistic (reshuffling of species among sites), and the suggestion that the species pool affects the null model is not unique (Shipley & Weiher, 1995). However, even sophisticated and complex null models need to be biologically relevant and interpretable, and null models are still frequently used incorrectly. Although only mentioned briefly, De Bello also notes another problem with studies of community assembly, which is that popular indices like FD, PD, and others may not always be able to distinguish correctly between different assembly mechanisms (Mouchet et al. 2010, Mayfield & Levine, 2010), something that null model do not control for. 

How is a species like a baseball player?

Biomass is to runs as species is to player, and as ecologist is to Brad Pitt.

Community ecology and major league baseball have a lot to learn from each other.

Let's back up. As a community ecologist, I think about how species assemble into communities, and the consequences for ecosystems when species disappear. I'm especially interested using traits of species to address these issues. For the grassland plants that I often work with, the traits are morphological (for example, plant height and leaf thickness), physiological (leaf nitrogen concentration, photosynthetic rate), and life history (timing and mode of reproduction).

As a baseball fan, I spend a lot of time watching baseball. Actually, I'm watching my Red Sox now (multitasking as usual; I freely admit there's a lot of down time in between pitches). I care about how the team does, mostly in terms of beating the Yankees. I'm especially interested in how individual players are doing at any time; for fielders I care about their batting average and defensive skills, and for the pitchers I care about how few runs they allow and how many strikeouts they get.

So my vocation and avocation have some similarities. Both ecology and baseball have changed in the last decade or so to become more focused on 'granular' data at the individual level. In ecology this has been touted as a revolutionary shift in perspective, but is really a return to the important aspects of what roles organisms play in ecosystems, and how ecosystems are shaped by the organisms in them. This trait-based approach has shifted the collection and sharing of data on organism morphology, physiology, and life history into warp speed, to the great benefit of quantitatively-minded ecologists everywhere.

In baseball, the ability to collate and analyze data on every pitch and every play has lead to an explosion of new metrics to evaluate players. One of the simplest of these new metrics, which even the traditionalists in baseball now value, is "on base plus slugging" (OPS, see all the details here). This data-intensive approach to analyzing player performance was most famously championed by the manager of the Oakland Athletics in the late 1990's, now being played by Brad Pitt in the upcoming movie Moneyball.

There is no one ecologist in particular who can claim credit for popularizing trait-based approaches in community ecology, but for the sake of laughs let's make Owen Petchey the Brad Pitt analogue.

What can we do with this analogy? For pure nerd fun, we can think about what these two worlds can learn from each other.

What can baseball learn from community ecology?

One of the most notable trait-centric innovations in community ecology has been the use of functional diversity (FD), which represents how varied the species in a community are in terms of their functional traits. Many flavors of FD exist (one of which was authored by Owen Petchey, above), but the goal is to use one value to summarize the variation in functional traits of species in a community. A high value for a set of communities indicates greater distinctiveness among the community members, and is taken to represent greater niche complementarity.

For fun, I've taken stats from a fantastic baseball database[i] and calculated the FD of all baseball teams from 1871 to 2010. I used a select set of batting, fielding, and pitching statistics[ii], and you can see the data here. For the two teams that I pay the most attention to, I plotted their FD against wins, with World Series victories highlighted:

Given that these FD values represent how different the members of a team are, it's surprising that there is much of a pattern at all. But the negative relationship between wins and FD is strong and significant by several measures[iii]. So: the more similar a team is in terms of player statistics, the better the team does!

This pattern of less dissimilarity among players correlating with better performance at the team level has apparently been noticed before, by Stephen Jay Gould, who extrapolated this pattern also across teams to explain the gradual shrinking of differences among players over time:

"if general play has improved, with less variation among a group of consistently better payers, then disparity among teams should also decrease"

and so:

"As play improves and bell curves march towards right walls, variation must shrink at the right tail." (from "Full House", thanks to Marc for this quote!).

Interesting, but is it useful? One obvious drawback in this approach of examining variation in individual performance is that it ignores the fact that in baseball, we know that a high number of earned runs allowed is bad for a pitcher, and a low number for hits is bad for a hitter. In contrast, a high value for specific leaf area is neither good nor bad for a plant, just an indication of its nutrient acquisition strategy.

There are many exponentially more nerdy avenues to go with applying community ecology tools to baseball data, but I'll spare you from that for now!

What can community ecology learn from baseball?

One new baseball stat that gets a lot of attention during trades is 'wins above replacement'. This is such a complicated statistic to calculate that the "simple" definition is that for fielders, you add together wRAA and UZR, while for pitchers it is based off of FIP. I hope that cleared things up.

The point in the end is to say how many wins a player is worth, when compared to the average player. In ecology, the concept of 'wins above replacement' has at least two analogies.

First, community ecologists have been doing competition experiments since the dawn of time. The goal is to figure out what the effect of a species is at the community level, although fully factorial competition experiments at the community level are challenging to carry out. For example, Weigelt and colleagues showed that there can be non-additive effects of competitor plant species on a target species, but could rank the effect of competitors. This result allowed them to predict the effect of adding or removing a competitor species from a mixture, in a roughly similar way to how a general manager would want to know how a trade would change his or her team's performance.

Second, ecologists have shown that both niche complementarity and a 'sampling effect' are responsible for driving the positive relationship between biodiversity and ecosystem functioning. The sampling effect refers to the increasing chance of including a particularly influential species when the number of species increases. Large-scale experiments in grasslands have been carried out where plants are grown in monoculture and then many combinations, up to 60 species. The use of the monocultures allows an analysis similar in spirit to 'wins above replacement', by testing how much the presence of a particular species, versus the number of species, alters the community performance.

We could take this analogy further, and think of communities more like teams. A restoration ecologist might calculate 'wins above replacement' for all the species in a set of communities, and then create All Star communities from the top performers.

Lessons learned

A. Shockingly, there are baseball nerds, and there are ecology nerds, and there are even double-whammy basebology nerds.

B. There are quantitative approaches to analyzing individual performance in these crazily disparate realms which might be useful to each other.

C. I might need to spend more time writing papers and less time geeking out about baseball!

More analogies to consider:

Reciprocal transplants: trades?

Trophic levels: minor league system?

Nitrogen fertilization: steroids?

---

[i] One of the most astonishing databases around: complete downloadable stats for every player since 1871. This database is what NEON should aspire to be, except that this one was compiled completely privately by some single-minded and visionary baseball geeks!

[ii] Batting: Hits, at bats, runs batted in, stolen bases, walks, home runs

Fielding: Put outs, assists, errors, zone rating

Pitching: Earned run average, home runs allowed, walks, strike outs.

[iii] E.g. even after taking into account other more typical measures of success in offense (runs, R) and defense (runs allowed, RA), within years, there is still a negative slope for FD on wins:

lme(win ~ R + RA + FD, random = ~1|yearID, data = team)

Value Std.Err DF t-value p-value

(Intercept) 80.289 0.7411 2159 108.3 <0.001

R 0.107 0.0009 2159 116.8 <0.001

RA -0.105 0.0009 2159 -115.6 <0.001

FD -1.729 0.8083 2159 -2.1 0.0325