This repository contains the R-based analytical pipeline for assessing how global climate and human land-use impacts shape the structural properties of ecological networks. The workflow processes mutualistic (plant-pollinator, seed-dispersal) and antagonistic (host-parasite) networks, integrating temporal and spatial environmental data to model changes in connectance, nestedness, modularity, and specialization.

To robustly account for spatial autocorrelation in these global datasets, the modeling framework utilizes Spatial Eigenvector Mapping (SEVM / dbMEM) and validates residuals using Monte Carlo Moran's I tests.

Scripts and analysis

This folder contains R scripts used to extract data, compute metrics, merge datasets, and produce figures and models for the project.

• r.Rproj
  • RStudio project file for convenient development. Open this file in RStudio to set working directory and project environment.

Network metrics

• metrics_antagonistic_1.R, metrics_antagonistic_2.R, metrics_antagonistic_3.R
  • Compute observed network-level metrics (connectance, modularity, nestedness, specialization) and generate null models (vaznull) for antagonistic networks. Split into 3 batches for parallel HPC processing.
• metrics_mutualistic_1.R ... metrics_mutualistic_7.R
  • Compute observed network-level metrics and generate null models for mutualistic networks (plant-pollinator, seed-dispersal). Split into 7 batches for parallel HPC processing.

Temporal matching

• temporal_match.R
  • Performs specific temporal matching of climate and human impact data to the sampling or publication year of each network. It automatically downloads and resamples TerraClimate netCDF files (precipitation, minimum/maximum temperature) to compute annual means and seasonality. It also extracts human population, built surface area (via GHS rasters), and the Human Footprint Index (HFI) across defined multi-year windows.

Models

• models_functions_REV.R
  • Contains the underlying helper functions for the spatial analysis. It includes coordinate projection transformations, custom functions to compute PCA-based climate effects, spatial weight builders (KNN-based distance), and functions to run Monte Carlo Moran's I tests and generate spatial correlograms. It also handles formatted HTML table outputs for model coefficients.
• models_SEVM.R
  • The primary analysis script. It prepares the data, runs a global climate PCA, and executes a customized modeling pipeline using distance-based Moran's Eigenvector Maps (dbMEM). It utilizes automated forward selection to identify significant spatial structures and fits linear mixed-effects models (lme) to evaluate environmental predictors while controlling for spatial autocorrelation and author random effects.

Figures

• figure1.R
  • Generates a global map of network sampling locations mapped by interaction type, alongside bivariate 2D density panels showing the distribution of networks across Mean Temperature and Human Footprint Index gradients.
• figure2_3.R
  • Creates unified PCA biplots of global climate variables and maps partial effect predictions. It builds multi-panel layouts displaying the marginal effects of climate (PC1/PC2) and human impacts (Population/HFI) on network metrics.
• figure4.R
  • Produces partial residual plots comparing the distinct responses of different taxa within the same interaction types (e.g., Birds vs. Mammals in seed-dispersal; Birds vs. Insects in pollination) across human impact gradients.
• figure_s1.R
  • Generates supplementary density plot (Fig. S1) for the network sampling years across the three interaction types.
• figure_s2.R
  • Generates supplementary density plot (Fig. S2) for the z-score values for specialization, modularity, and nestedness across the three interaction types.

Dataframe for analysis

• _weighted_antagonistic_analysis.csv
  • The consolidated dataset for antagonistic networks, containing all calculated network metrics, null-model standardizations (z-scores), and matched spatial/environmental variables.
• _weighted_mutualistic_analysis.csv
  • The consolidated dataset for mutualistic networks, containing all calculated network metrics, null-model standardizations (z-scores), and matched spatial/environmental variables.

Variable descriptions

The following section describes the variables (column headers) included in the two analytical datasets: _weighted_antagonistic_analysis.csv and _weighted_mutualistic_analysis.csv.

• ID: Unique identifier for the specific network dataset.
• Web_Type: The ecological interaction type of the network (e.g., host_parasite, plant_pollinator, seed_dispersal).
• Author: The author group used as a random effect.
• taxa: Broad taxonomic group(s) involved in the interactions (e.g., Insects, Birds, Fish, Mammals).
• year_network: The year the network interaction data was collected in the field.
• year_publication: The year the study describing the network was published.

Network metrics:

• Network_Size: Total number of interacting species within the network (sum of both levels).
• Connectance: The realized proportion of all possible interactions in the network. Ranges from 0 to 1.
• Quantitative_Specialization: The observed weighted specialization of the network (e.g., H₂').
• Weighted_Nestedness: The observed weighted nestedness of the network (e.g., WNODF).
• Weighted_Modularity: The observed weighted modularity of the network (e.g., Q).

Null-model standardizations:

• *_zscore (e.g., Quantitative_Specialization_zscore, Weighted_Nestedness_zscore, Weighted_Modularity_zscore): The standardized effect size (z-score) of the observed metric relative to the null models.
• Mean_* (e.g., Mean_Weighted_Modularity): The average value of the specific metric calculated across all generated null model networks.
• SD_* (e.g., SD_Weighted_Modularity): The standard deviation of the specific metric across all generated null model networks.

Spatial and environmental variables:

• lat: Latitude of the study site (Decimal Degrees).
• long: Longitude of the study site (Decimal Degrees).
• t_mean: Mean annual temperature (°C).
• t_seasonality: Temperature seasonality (Standard deviation of temperature * 100).
• ppt_mean: Mean annual precipitation (mm).
• ppt_seasonality: Precipitation seasonality (Coefficient of variation).
• Built_Surface: Proportion or measure of built-up surface area matching the site location.
• Human_Population: Estimated human population density/count at or near the site coordinates.
• hii_matched: Matched Human Footprint Index (HFI) score, representing the cumulative anthropogenic impact on the landscape.

Only in _weighted_antagonistic_analysis.csv:

• .row_id: An internal row identifier generated during data processing.
• Num_Hosts: Total number of host species present in the antagonistic network.
• Num_Parasites: Total number of parasite species present in the antagonistic network.

Only in _weighted_mutualistic_analysis.csv:

• Num_Plants: Total number of plant species present in the mutualistic network.
• Num_Pollinators_or_dispersers: Total number of animal species (pollinators or seed dispersers) present in the mutualistic network.