(Crossposted with edits from the Ecography Blog; original post July 8, 2014)
In July 2014, we (my collaborator Jeremy Lundholm, our joint PhD student Oluwatobe “Tobi” Oke, and I) published a paper in Ecography: “Integrating phylogenetic community structure with species distribution models: an example with plants of rock barrens”. (And kudos to Holly Abbandonato for 1st-rate field help). I wrote the following “story behind the paper” for Ecography’s blog. I like reading this kind of thing, so you’ll probably see more on the blog in future.
Our paper combines approaches from phylogenetic community ecology and species distribution modeling to understand the assembly of plant communities on rock barrens. It was enormous fun to be involved with the work, in part because before we started I knew nothing about SDMs and next to nothing about rock barrens. That we ended up with what I think is a pretty good paper is a testament to the value of collaboration and coauthorship.
The paper evolved from Tobi’s graduate work. He initially wanted to ask whether phylogenetic community structure could predict ecosystem function in green roofs. (A ‘green roof’ uses plants grown on a building’s roof to help manage rainwater runoff and to provide other ecosystem services). I knew something about phylogenetic community structure, and Jeremy knew a lot about green roofs. This made it an interesting project for both of us, and we saw potential for our complementary experience to let us connect those subfields. But it was the unplanned connections as Tobi’s project developed that I think really paid off. Here’s how that played out.
Rock barrens (treeless areas within forests where bedrock lies at or just below the soil surface) are a natural analogue to green roofs, which minimize weight by using plants grown in shallow trays with relatively little soil. In regions where barrens occur, they’re also a potential source of native plant species for deployment in green roofs. Tobi intended to study the green-roof performance of barrens plants grown in different combinations. As a point of comparison, he decided to measure phylogenetic community structure on the natural barrens from which he was collecting. This was mostly straightforward (although I should have realized more quickly that sites defined by the absence of trees wouldn’t be well-served by New Brunswick’s network of logging roads!).
At the quadrat scale, our barrens plants showed strong phylogenetic clustering, a pattern that’s often presumed to arise from environmental filtering – that is from community membership being determined by tolerance of local environmental conditions, rather than by the outcome of competition. This makes sense if closely related species have similar environmental tolerances: each quadrat will contain a set of close relatives that share tolerance for conditions there. But phylogenetic community ecology has recently seen some pointed discussion over whether it’s wise, or even possible, to infer processes behind community assembly from patterns of relatedness among co-occurring species. So far, we had no way to close the gap between pattern and process.
The most obvious way to close the pattern-process gap would have been to gather ecological trait data for all our plant species, so that we could ask whether the traits determining occurrence at a site are phylogenetically conserved (so that filtering could drive phylogenetic clustering). But this is a daunting prospect: we’d have to identify the abiotic factors determining local occurrence, figure out which plant traits were relevant to those abiotic factors, and then measure trait values for all our species. Very few studies have managed to do all that, and I doubt that we could have either.
Tobi, though, had a really exciting insight: instead, maybe we could close the gap using species distribution modeling. As he explained this (because I was a complete rookie), I got the feeling we were about to make a real leap. After all, SDMs can be built from plant occurrence data that are easily available, and can be used to deduce the abiotic factors controlling plant distributions. And although nobody seemed to have taken advantage of the fact, the parameter values from fitted SDMs can be interpreted as trait values describing the response of each plant species to those abiotic factors! They aren’t “conventional” functional traits like specific leaf area or stomatal conductance, but in a way they’re even better: they’re ecologically integrative, measuring the overall response of the plants to each modeled abiotic factor. If we could pull this off for our barrens species, we’d have the trait values we needed to connect phylogenetic pattern to ecological process.
We think we did pull it off. For our 9 most common barrens species, SDMs strongly suggested drought stress as the most important abiotic driver of plant distribution. The parameter values measuring the shapes of our species’ occurrence functions, which we interpret as drought-tolerance traits, showed strong phylogenetic pattern. Together, these facts support a process-based explanation (environmental filtering on drought tolerance) for the observed pattern in phylogenetic community structure (strongly phylogenetic clustering). Fist pumps were in order!
What we don’t know yet, of course, is whether our approach will work for other communities the way it worked for us in rock barrens. Maybe it won’t; but if it does, it promises a powerful approach to studying community assembly based on data that are relatively simple to obtain: distributional data from herbarium records, local co-occurrence data from quadrat surveys, and phylogenetic hypotheses from online tools that are (mostly) automated and easy to use. All this because Tobi realized that two apparently unrelated techniques, twisted just the right way, fit together to do something new.
Finally, I think there is a useful lesson to be had for our discipline as a whole. In the very first post on my blog, I talked about some advantages and disadvantages of a research program that frequently hops among fields. While the disadvantages are real, one advantage is that field-hopping can make it easier to spot synergies between ideas developed in different literatures. We think our barrens paper is a good example of such unexpected synergy, and we were all very excited to see it develop.