MultilayerEcoNet

GitHub repository for the manuscript On the structure of species-function participation in multilayer ecological networks and its Supplementary Information. Here, you'll find all the necessary materials to reproduce all results contained in the manuscript https://www.nature.com/articles/s41467-024-53001-1, including input data stored in the 'data/input' folder and output results stored in the 'data/output' folder.

P processing

All computational analysis was made using Python 3.10.8 running in a Linux desktop (version 22). Dependencies are listed at the beginning of each file. python3_enviroment contains the requirements.txt and enviroment_droplet.yml files needed to recreate a python environment with all libraries adn dependencies needed. To install the environment with conda, run conda env create -f environment_droplet.yml.

1- P_processing is a folder that contains the necessary material to reproduce most of the results shown in the main manuscript and in the Suplementary Information.

2- FullCode.ipynb is the notebook, and functions used are stored in FullCodeFunctions.py.

3- FullCode18vs25.ipynb is the notebook to compare data from different seasons/years in an ecosystem, and functions used are stored in FullCodeFunctions.py.

4- Two very resource consuming plots have been separated from the rest. They are in the folder MFENplot, wchich contains two python scripts to depict the Multifunctional Ecological Network (MFEN) (Print_MultiGraphs.ipynb generates Figure 2 from the main manusccripy and Print_MultiGraphs_info_multi.ipynb Figure 1a from SI) and a third one with the functions used in both (Plot_Mcomm_Lib.py) using Netgraph (https://github.com/paulbrodersen/netgraph).

P analytics

In P_analytics.ipynb:

1- Import the resource-function matrix $\mathbf{P}$, computed from MFEN of the Na Redona dataset based on Eq.1 of the paper generated in FullCode.ipynb.

2- Visualize $\bf P$ along with some randomizations and compute NODF for binarizations.

3- Build nestedness-based rankings (ordering species and functions in terms of their participation strength in $\mathbf{P}$), along with suitably defined null models.

4- Build and plot $\mathbf{\Phi}$ and $\mathbf{\Pi}$.

5- Look into hierarchical clustering properties of $\mathbf{P}$.

6- Compute the conditioned $\mathbf{\Phi}|_i$ and $\mathbf{\Pi}|^\alpha$, and the keystonness scores.

7- Build a null model for the keystonness scores.

8- Functions used are stored in FullCodeFunctions.py.

Nestedness

NODF_WNODF contains all the necessary material to reproduce NODF and WNODF results and null models from the main text using fortran90 using the resource-function matrix $\mathbf{P}$ of the paper generated in FullCode.ipynb.

For NODF compile using:

f95 -O3 read_n.f nest_nodf.f dranxor.f90 -o nest_n.x

and run using:

./nest_n.x

For WNODF compile using:

f95 -O3 read_w.f nest_wnodf.f dranxor.f90 -o nest_w.x

and run using:

./nest_w.x

The random number generator used is: https://ifisc.uib-csic.es//raul/CURSOS/Stochastic_Simulation_Methods/dranxor.f90

ExecuteWNODFPython.ipynb allows to execute fortran code for NODF and WNODF from Python

Authors

Study site, Field sampling and Data curation:

• Sandra Hervı́as-Parejo
• Anna Traveset
• Isabel Donoso
• Ruben Heleno
• Manuel Nogales
• Susana Rodrı́guez-Echeverrı́a

Data analysis, Mathematical modelling, Network analysis and Simulations:

• Mar Cuevas-Blanco (P processing, P analytics and MFENplot)
• Lucas Lacasa (P analytics)
• Victor M. Eguiluz (NODF_WNODF)
• Carlos J. Melian (provided guidance and advice)

Feel free to reach out to me anytime! 💫