Dataset DOI: 10.5061/dryad.70rxwdc8c

Description of the data and file structure

This dataset was collected as part of a 1.5-year field study examining how intraspecific variation and seasonal dynamics shape seed dispersal networks in gopher tortoises (Gopherus polyphemus) inhabiting South Florida’s pine rockland ecosystem. Fourteen radio-tracked tortoises were monitored via telemetry across three spatially distinct sites in the Richmond Tract of Miami-Dade County, Florida. Fecal samples were opportunistically collected and dissected to identify seeds consumed and dispersed by each individual tortoise. For each sample, the date, tortoise ID, and GPS location were recorded. Identified seeds were categorized by plant species and dispersal syndrome (e.g., endozoochorous vs. other). Morphological traits—including straight-line carapace length, carapace width, gape size, and body mass—were measured for each tortoise. These data were used to construct individual-based annual and seasonal bipartite seed dispersal networks, assess network-level and node-level properties (e.g., modularity, specialization, partner diversity), and model the influence of tortoise morphology and seasonal fruiting phenology on interaction structure. The dataset underpins analyses presented in the associated manuscript published in Oikos.

Files and variables

File: Figueroa_et_al._2025_Oikos_Data.xlsx

Description:

This Excel workbook contains data collected from 14 radio-tracked gopher tortoises (Gopherus polyphemus) in the South Florida pine rockland ecosystem over a 1.5-year period. Fecal samples were dissected to identify ingested seeds and used to construct individual-based annual and seasonal bipartite seed dispersal networks. Morphological traits were also measured for each individual. This dataset underlies all network, seasonal, and trait-based analyses presented in the associated Oikos manuscript.

Variables

• In the sheets "All", "Dry", "Wet", and "All.Modules", the column "species" contains the scientific name of each plant species detected in at least one fecal sample from the tortoises.

• In the sheets "Dry.syndrome" and "Wet.syndrome", the column "syndrome" indicates the dispersal syndrome of each plant species detected in tortoise diets. Categories include: anemochory, autochory, ballistochory, ectozoochory, endozoochory, foliage is fruit, and myrmecochory, based on classifications from Ridley (1930), Janzen (1984), and Van der Pijl (1982).

• In all sheets listed above, all columns (except the first column) contain the proportion of fecal samples from each tortoise (identified by ID in the column header) that included seeds of that plant species or syndrome. Values range from 0 to 1.

• In "plant.trait.analysis", the column "syndrome" is classified as either "endozoochory" (i.e., seeds enveloped in fleshy fruit) or "other".

• In "plant.trait.analysis" and "seasonal.trait.analysis", the column "season" denotes when the sample was collected and is categorized as either "Dry" or "Wet".

• In "all.trait.analysis", the column "samples" indicates the total number of samples collected for each tortoise individual.

• In all sheets where they appear:

 – "ind" is the unique tortoise identifier

 – "site" denotes the site where the tortoise was located: E (East), S (South), or W (West)

 – "cl" and "cw" are straight-line carapace length and carapace width, respectively, measured in centimeters (cm)

 – "gape" is the gape width measured in millimeters (mm)

 – "mass" is the average body mass in grams (g)

 – "sex" indicates adult sex as "M" (male) or "F" (female)

 – "age.class" categorizes individuals as "M" (adult male), "F" (adult female), or "J" (juvenile of undetermined sex)

Network Metrics (produced via specieslevel() in the bipartite R package)

• d (Blüthgen’s d′): A measure of interaction specialization that quantifies how selectively a node interacts relative to the availability of potential partners.

• Values range from 0 (completely generalized) to 1 (completely specialized).
• Accounts for partner availability and interaction frequency.

• norm.deg (Normalized degree): The proportion of all possible plant species that a tortoise interacts with.

• Calculated as: (number of partners used) / (total number of available partners)
• Values range from 0 to 1. Does not account for interaction strength.

• partner.diversity: A Shannon diversity index reflecting the diversity and evenness of plant species interacted with.

• Calculated as: -Σ (p_i * log(p_i)) where p_i is the proportion of interactions with plant species i.
• Higher values indicate more evenly distributed interactions.

• species.strength: Sum of the dependency values that all partners place on a given node.

• Interpreted as the importance of the node (e.g., a tortoise or plant) in the network.
• Higher values indicate greater centrality or influence in the interaction web.

• In "all.trait.analysis", the column "module" indicates the modular group each tortoise was assigned to based on the computeModules() function in the bipartite package.

Missing Values:

No special indicators such as "NA" or blank cells are used. Absence of interaction is coded as 0.

Literature Cited:

Janzen, D. H. 1984. Dispersal of Small Seeds by Big Herbivores: Foliage is the Fruit. The American Naturalist 123:338–353.

Ridley, H. N. 1930. The Dispersal of Plants Throughout the World. L. Reeve & Company, Limited.

Van der Pijl, L. 1982. Principles of Dispersal in Higher Plants. Springer.

Code/software

The dataset was analyzed using R version 3.3.0, an open-source statistical computing environment. Several free R packages were used to process the data, calculate individual- and network-level metrics, and generate the outputs reported in the associated manuscript.

Required software and packages:

• R (v3.3.0) – https://www.r-project.org
• Packages:
  • bipartite – for constructing bipartite seed dispersal networks and calculating network and species-level indices (e.g., specieslevel(), networklevel(), computeModules()).
  • readxl – for importing Excel data.
  • tidyverse – for data wrangling (dplyr, tidyr) and reshaping.
  • betalinkr – for quantifying temporal change in individual-based networks using beta diversity metrics (beta.temp()).
  • ggplot2 – for data visualization.

Workflow summary:

1. Data import and cleaning: Seed interaction matrices were imported from .xlsx files using readxl, and reshaped using tidyverse tools.
2. Network construction and analysis:
   • Annual and seasonal bipartite matrices were analyzed using bipartite::specieslevel() and bipartite::networklevel() to calculate individual-level metrics (e.g., d', normalized degree, species strength, partner diversity).
   • Modularity structure was assessed using bipartite::computeModules().
3. Trait analyses: Morphological traits of tortoises were merged with network metrics and analyzed to identify trait-network correlations.
4. Temporal comparison of networks: The betalinkr package was used to quantify turnover in network composition and structure between wet and dry seasons using beta.temp().

Code Summary:

This R script: Oikos_GT_Network_Code.R conducts a comprehensive analysis of seed dispersal networks constructed from fecal samples of radio-tracked gopher tortoises (Gopherus polyphemus) inhabiting the pine rocklands of South Florida. The workflow includes:

1. Data Import and Preparation

• Fecal sample data were imported from .xlsx files using readxl and converted into bipartite matrices representing individual tortoises (rows) and plant species or syndromes (columns).
• Annual, seasonal (dry/wet), and syndrome-level matrices were prepared and converted to integer-weighted matrices for downstream network analysis.
• Color-coded web visualizations were generated using bipartite::plotweb() to depict interaction structure across seasons and individuals.

2. Network Metrics and Visualization

• Node-level metrics including Blüthgen’s d′, normalized degree, species strength, and partner diversity were calculated using bipartite::specieslevel() for each matrix (annual and seasonal).
• Network-level metrics such as modularity, nestedness (wNODF), specialization (H2′), interaction evenness, linkage density, and weighted connectance were quantified via bipartite::networklevel().
• Null model comparisons (1,000 iterations using vaznull) were used to assess significance of observed network metrics.

3. Modularity and Community Structure

• Module affiliation was calculated using bipartite::computeModules() and visualized with plotModuleWeb() for the overall and seasonal networks.
• Individuals were assigned to modular groups and analyzed for trait-level correlates.

4. Temporal Dissimilarity in Interactions

• Seasonal turnover in interaction structure was quantified using betalinkr::betalinkr() to partition dissimilarity into species turnover (β_ST) and shared species rewiring (β_OS).

5. Trait-Based Bayesian Modeling

• Generalized linear multilevel models (GLMMs) were fitted using brms (with cmdstanr backend) to test the influence of tortoise morphological traits (e.g., carapace length, width, gape size, mass) on d′ and partner diversity.
• Seasonal effects on specialization metrics were tested for both tortoise and plant communities.
• Interaction plots and marginal effects were generated with tidybayes, marginaleffects, and ggplot2.

6. Null Model Comparisons at Node-Level

• Confidence intervals (95% CI) and standard deviations for each node-level metric were calculated from null distributions for each individual or species.

These null values were extracted and organized by node identity to statistically compare observed versus expected specialization patterns.

Access information

This dataset is original and was not derived from any other publicly accessible source. It is not available elsewhere and is exclusively archived on Dryad. No external licenses apply.