Data and code from: Network theory predicts ecosystem robustness across environmental conditions
Data files
Jul 02, 2025 version files 2.05 MB
Abstract
Network theory quantifies how changes in species richness, S, lead to changes in the number of interactions (or links) between species, L. Networks with a steep relationship between L and S have a high number of links per species, making the network resistant to collapse and therefore more robust. However, changes in S often coincide with environmental shifts, which can lead to impacts on L that are not expected from network theory. In this paper, we constructed relationships between L and S for 18 ecosystems using 1081 observations collected across 420 environmental conditions. We found that environmental noise (unspecified spatiotemporal variation) and environmental gradients (directional environmental change) commonly affected ecological network size (S and L), community composition, and also induced network rewiring, which means that species changed interaction partners as the environment changed. Yet, we found the log(L) ~ log(S) relationship to be remarkably constant across environmental conditions. Specifically, the slope of this relationship remained constant across conditions, implying consistency in how species loss proportionally affects L. Our results therefore show that network theory predicts ecosystem robustness across environmental conditions. These results suggest generality to how environmental drivers operate at the level of ecological networks, which is encouraging for conservation.
Title: Network theory predicts ecosystem robustness across environmental conditions
Authors: Germain Agazzi, Camille Carpentier, Olivia Bleeckx, and Frédérik De Laender
Contact: germain.ag@gmail.com or frederik.delaender@unamur.be
Introduction
In this paper, we reused datasets from previously published papers in the literature. The zip file "Datasets_NetworkPredictRobustnessEnvironmentalConditions" contains these datasets, which have been transformed from their original to fit to the format used in this study. All of these datasets are described in more detail in their original publications (references are in the following text).
The "dataset" section summarizes them, and the "code" section explains the code.
Dataset
All datasets contain common columns:
- A and B: names of the two interacting species (sometimes represented by codes)
- inter: name of the interaction between A and B
- category: category of the network (antagonistic or mutualistic)
- partition: partition type of the network (unipartite or bipartite)
All of them also contain other columns that are detailed below.
File Alcantara (from Julio M. Alcántara and Pedro J. Rey. Linking Topological Structure and Dynamics in Ecological Networks. The American Naturalist, August 2012):
- web: code for the web (location_code)
- location: location name
- code: treatment code
File Chiu (from Chun-Huo Chiu, Anne Chao, Sebastian Vogel, Peter Kriegel, and Simon Thorn. Quantifying and estimating ecological network diversity based on incomplete sampling data. Philosophical Transactions of the Royal Society B: Biological Sciences, 378(1881):20220183, May 2023):
- plot: code referring to the plot
- treatment: treatment abbreviation, "G" for naturally shaded, "N" for artificially shaded, and "O" for open
- year: sampling year
File David (from Jeanne David and Renate A. Wesselingh. Étude de la biodiversité des espaces vert gérés par l’UCLouvain sur le campus de Louvain-la-Neuve: type de gestion et réseaux plantes-pollinisateurs. Master thesis, Catholic university of Louvain La Neuve, 2023):
- site: site name
- gestion: gestion method, "Tonte" for mowing and "Fauche" for reaping
- date: sampling month, 'Juin" for June "miJuillet" for "mid July" and "finJuillet" for "end July"
File Guardiola (from Moisès Guardiola, Constanti Stefanescu, Ferran Roda, and Joan Pino. Do asynchronies in extinction debt affect the structure of trophic networks? A case study of antagonistic butterfly larvaeplant networks - Guardiola - 2018 - Oikos - Wiley Online Library. Oikos, November 2017):
- web: Id of the web
File Kemp (from Jurene E. Kemp, Darren M. Evans, Willem J. Augustyn, and Allan G. Ellis. Invariant antagonistic network structure despite high spatial and temporal turnover of interactions. Ecography, 40(11):1315–1324, 2017):
- location: location name
- date: season
File Nielsen (from Anders Nielsen and Ørjan Totland. Structural properties of mutualistic networks withstand habitat degradation while species functional roles might change. Oikos, 123(3):323–333, 2014):
- site: site code
- web: ID web (site_state)
- state: gestion state, "CC" for clear cut, "YF" for young forest, and "OGF" for old growth forest
File Olito (from Colin Olito and Jeremy W. Fox. Species traits and abundances predict metrics of plant–pollinator network structure, but not pairwise interactions. Oikos, 124(4):428–436, 2015.):
- tsect: transect code
- jdate: sampling day
- meter: altitude
File Saavedra(from Serguei Saavedra, Simone Cenci, Ek Del-Kal, Karina Boege, and Rudolf P. Rohr. Reorganization of interaction networks modulates the persistence of species in late successional stages. Journal of Animal Ecology, May 2017):
- state: successional state
- location: location code
- date: date code
File Saravia (from Leonardo A. Saravia, Tomás I. Marina, Nadiah P. Kristensen, Marleen De Troch, and Fernando R. Momo. Ecological network assembly: How the regional metaweb influences local food webs. Journal of Animal Ecology, 91(3):630-642, 2022):
- foodweb.name: web name
- study.site: ecosystem name
File VanderZee (from Els M. van der Zee, Christine Angelini, Laura L. Govers, Marjolijn J. A. Christianen, Andrew H. Altieri, Karin J. van der Reijden, Brian R. Silliman, Johan van de Koppel, Matthijs van der Geest, Jan A. van Gils, Henk W. van der Veer, Theunis Piersma, Peter C. de Ruiter, Han Olff, and Tjisse van der Heide. How habitat-modifying organisms structure the food web of two coastal ecosystems. Proceedings of the Royal Society B: Biological Sciences, 283(1826):20152326, March 2016):
- site: site code, "BdA" for Africa and "RI" for America
- treatment: treatment code, "B" for bare ground, "C" for cobblestone, and "E" for colonized
- plot: plot code
File Welti (from Ellen A.R. Welti, Fan Qiu, Hannah M. Tetreault, Mark Ungener, John Blair, and Anthony Joern. Fire, grazing, and climate shape plant–grasshopper interactions in a tallgrass prairie. Functional Ecology, December 2018.):
- year: sampling year
- plot: plot code
File Wood (from Spencer A. Wood, Roly Russell, Dieta Hanson, Richard J. Williams, and Jennifer A. Dunne. Effects of spatial scale of sampling on food web structure. Ecology and Evolution, 5(17):3769–3782, September 2015.):
- WebID: web ID
- WebScale: sampling scale, "Q" for quadrats, "L" for locales and "S" for sites
Code
The code (Code_NetworkPredictRobustnessEnvironmentalConditions.R) is run under R 4.4.2 with the Following packages:
- reshape2 1.4.4
- tidyverse 2.0.0
- car 3.1-3
- arm 1.14-4
- rstatix 0.7.2
- fitdistrplus 1.2-1
- lme4 1.1-35.5
- ade4 1.7-22
- factoextra 1.0.7
- lmerTest 3.1-3
- MuMIn 1.48.4
- performance 0.12.4
- stringr 1.5.1
- vegan 2.6-8
- ggpmisc 0.6.1
- cowplot 1.1.3
- patchwork 1.3.0
- ggpubr 0.6.0
The code is divided into several parts:
- setup: setup, importations, commands, and packages
- analyses: individual analyses for each dataset
- all: merging of all datasets and analyses
- graphs: code for generating figures
- Diff: beta diversity analyses
We searched the literature to collect ecological network datasets with at least eight observations per ecosystem (to allow for meaningful linear regressions), ensuring that these datasets were based on field observations rather than simulations. Network observations were coded as undirected adjacency matrices, where 0 and 1 indicated the absence or presence of an interaction, respectively. For each network observation, we recorded the number of species (S), the number of interactions (or links, L), and the composition of observed species and links. In cases where organisms were not identified at the species level, the term "species" referred to taxa.