Emergent shapes of trait-based competition functions from resource-based models

This repository contains all the code used in the manuscript:

Falster DS, Kunstler G, FitzJohn RG, Westoby M (2021) "Emergent shapes of trait-based competition functions from resource-based models: A Gaussian is not normal in plant communities." The American Naturalist

Abstract: In community ecology, it is widely assumed that organisms with similar traits compete more intensely with one another for resources. This assumption is often encoded into theory and empirical tests, via a unimodal competition function, which predicts that per capita competitive effect declines with separation in traits. Yet, it remains unknown how well this function represents the true effect of traits on competitive outcomes, especially for long-lived plant communities where lifetime fitness is difficult to estimate. Here we evaluate the shape of competition functions embedded in two resource-based (RB) models, wherein plants compete for shared, essential resources. In the first RB model, individuals compete for two essential nutrients, and in the second they compete for light in a size-based successional setting. We compared the shapes of the competition functions that emerged from interactions within these RB models to the unimodal function and others shapes commonly applied. In few instances did the trait-based competition function emerging from the RB model even vaguely resemble any of the shapes previously used. The mismatch between these two approaches suggests that theory derived using fixed competition functions based on trait separation may not apply well to plant systems where individuals compete for shared resources. The more promising path will be to model depletion of resources by populations in relation to their traits, with its consequences for fitness landscapes and competitive exclusion.

Additional accessory files are also including:

• DECRIPTION: A machine-readable compendium file containing key metadata and dependencies
• LICENSE: License for the materials
• Dockerfile & .binder/Dockerfile: files used to generate docker containers for long-term reproducibility

Running the code

All analyses were done in R, and the paper is written in LaTeX. All code needed to reproduce the submitted products is included in this repository. To reproduce this paper, run the code contained in the analysis.R file. Figures will be output to a directory called output and the paper and supplementary materials in the folder ms.

The paper was written in 2016 using a version of R available at the time. With some minor updates, the code has been updated and was last seen running wild and free on R 4.0. You can try running it on your current version and it may work.

To ensure computational reproducibility into the future, we have also generated Docker and Binder containers, enabling you to launch a compute environment built off R 3.6.1 with all the dependencies included.

Running locally

If reproducing these results on your own machine, first download the code and then install the required packages, listed under Depends in the DESCRIPTION file. This can be achieved by opening the Rstudio project and running:

#install.packages("devtools")
devtools::install_deps()

Then run analysis.R.

Running on Binder

You can launch the analysis on the web in an interactive RStudio session with the required software pre-installed. This session is hosted by binder and can be accessed by clicking on the following:

Running via Docker

If you have Docker installed, you can recreate the compute environment as follows.

First fetch the container:

docker pull traitecoevo/competition_kernels

Then launch it via:

docker run --user root -v $(pwd):/home/rstudio/ -p 8787:8787 -e DISABLE_AUTH=true traitecoevo/competition_kernels

The code above initialises a docker container, which runs an rstudio session, which is accessed by pointing your browser to localhost:8787. For more instructions on running docker, see the info from rocker.

Note, this container does not contain the actual github repo, only the software environment. If you run the above command from within your downloaded repo, it will map the working directory as the current working directory inside the docker container.

Building the docker images (optional)

For posterity, the docker image was built off rocker/verse:3.6.1 container via the following command, in a terminal contained within the downloaded repo:

docker build -t traitecoevo/competition_kernels .

and was then pushed to dockerhub (here). The image used by binder builds off this container, adding extra features needed bi binder, as described in rocker/binder.

Contributors

Daniel Falster Rich FitzJohn Georges Kunstler Saras Windecker (docker)