Reconstructing food webs in deep time: why model choice matters for ecological inference

This repository contains the full analytical pipeline used to evaluate how food-web reconstruction models influence ecological inference in paleoecosystems. Using the Early Jurassic (Toarcian) Oceanic Anoxic Event as a case study, multiple network reconstruction approaches are compared to assess how their assumptions shape predictions of ecosystem structure, extinction dynamics, and recovery.

Quick overview

• Languages: Julia and R
• Primary analysis scripts: /code (numerically ordered)
• Notebooks and figures: /notebooks
• Data: /data (raw + processed; large processed files are not included)

Quickstart (recommended)

1. Install Julia (recommended version: 1.9+).

2. Open a Julia REPL in the repository root and run:

   using Pkg Pkg.activate("code") Pkg.instantiate()

3. (One-off) Install required GitHub-only packages:

   pkg> add https://github.com/BecksLab/pfim.jl pkg> add https://github.com/BecksLab/Extinctions.jl

4. Run the analysis (examples):

   julia --project=code code/01_build_networks.jl julia --project=code code/02_topology.jl julia --project=code code/03_generate_extinctions.jl

5. Render the manuscript and Supplementary Materials (Quarto):

   quarto render .

Notes:

• Scripts are designed to be run in order, but individual steps can be re-run as needed.
• Long-running jobs can be run on a cluster or backgrounded locally.

What the pipeline does

• Generates ensembles of synthetic food webs using multiple reconstruction models (ADBM, ATN, Body-mass ratio, Niche, Random, PFIM variants).
• Computes network metrics at macro, meso (motifs), and micro scales.
• Simulates primary extinction sequences and propagating secondary extinctions under multiple removal scenarios.
• Quantifies interaction beta-diversity and performs statistical comparisons (MANOVA, LDA, PERMANOVA, mixed models).

Project structure (high level)

• /code: analysis scripts (01_... to S2_...)
  • /code/lib: reusable functions and model implementations
• /data
  • /data/raw: curated input trait and guild data
  • /data/processed: generated networks and analysis outputs (not tracked when large)
• /notebooks: supplementary outputs and tables used in the manuscript
• Quarto files (.qmd) in the repository root for manuscript rendering

Reproducibility

• Use the Project.toml and Manifest.toml inside /code to reproduce the Julia environment exactly.
• The Manifest snapshots package versions; keep it intact to ensure identical analysis results.
• External packages hosted on GitHub (PFIM.jl and Extinctions.jl) must be installed separately (see Quickstart).

Data availability

Large processed files are not stored in the repo. To reproduce results, run the relevant Julia scripts to regenerate the following (examples):

• data/processed/extinction_seq.jlds

If you need access to the generated datasets, contact the repository maintainers or check any supplementary data archive associated with the manuscript.

Contributing

Contributions and bug reports are welcome. Please open issues or pull requests describing the change and the rationale.

Citation

If using results or code from this repository, please cite the associated manuscript (when available).

Contact

Maintainer: Tanya STrydom