Related article title: Seeding success: Integrating seed dispersal networks in tropical forest restoration
Authors: Nicole M. Lussier, Rachel E. Crafford, J. Leighton Reid, Charles Kwit
Corresponding author: Nicole M. Lussier (lussie12@msu.edu)
Related publication: DOI: 10.1111/btp.13347
File Name: Lussier.etal.2024.xlsx

Descriptions of data and files

This dataset represents a compiled database of seed dispersal interaction networks extracted from published studies conducted in tropical ecosystems. Each row in the dataset corresponds to a single ecological network, as reported in an original study. Studies that reported multiple networks (e.g., multiple sites, habitats, or sampling regimes) contribute multiple rows to the dataset.

Networks are uniquely identified using a combination of two variables:

• Comp_ID, which identifies the published study, and
• Net_ID, which identifies individual networks within a study.

Together, Comp_ID and Net_ID form a unique identifier for each network.

The dataset integrates bibliographic information, geographic context, taxonomic focus, field sampling methodology, and reported network-level metrics. All quantitative network metrics were extracted directly from the original publications. When necessary, reported values were standardized to ensure consistency across studies (for example, converting percentages to proportions).

No raw interaction matrices are included in this dataset. Instead, the dataset focuses on reported network summaries (e.g., richness, modularity, nestedness), allowing for cross-study comparisons of network structure, sampling methodology, and taxa.

This dataset is intended to support comparative, meta-analytic, and synthesis-based research on seed dispersal networks, particularly in the context of tropical forest ecology, conservation, and restoration.

---

Data Collection, Processing, and Usage Notes

Literature Survey and Study Selection

A comprehensive literature survey was conducted to identify studies focused on seed dispersal networks in tropical ecosystems. Searches were performed using Web of Science, PubMed, and Google Scholar, as well as the Interaction Web Database and the Web of Life Ecological Network Database. Search terms included combinations of “seed dispersal network,” “plant–frugivore network,” “mutualistic interaction network,” and related keywords.

Studies were included if they:

1. Explicitly focused on seed dispersal networks and reported or analyzed interaction networks,
2. Were conducted in tropical regions,
3. Were published in peer-reviewed journals or dissertation manuscripts,
4. Were published between 1990 and 2023, and
5. Used empirical or theoretical approaches to study seed dispersal interactions.

To avoid duplication, studies that reused previously published empirical networks (e.g., in reviews or meta-analyses) were excluded unless they were the original source of the network data. After full-text screening, a total of 77 studies were retained.

---

Network Metric Extraction and Standardization

From each included study, reported network-level metrics were extracted, including:

• Network, plant, and disperser richness,
• Interaction frequency or total number of interactions,
• Modularity, specialization (H2′), nestedness (NODF), weighted nestedness (WNODF),
• Connectance, robustness, interaction diversity, interaction evenness, and niche overlap,
• Sampling completeness when reported.

When metrics were reported using different scales or formats across studies, values were standardized for consistency (e.g., converting sampling completeness from percentages to proportions). Metrics were only included if they could be clearly interpreted and aligned with standard ecological network definitions.

Missing values indicate that a metric was not reported in the original publication. No values were inferred or estimated when metrics were absent.

---

Field Methods and Sampling Classifications

Each study was categorized based on its primary field sampling approach:

• Phytocentric methods emphasize plant-centered observations (e.g., focal tree observations),
• Zoocentric methods emphasize animal-centered sampling (e.g., mist-netting, fecal analysis),
• Mixed methods combine both approaches.

These classifications allow users to examine how sampling strategies influence reported network structure and metrics.

---

Usage Notes and Limitations

Users should be aware that:

• Network metrics were calculated by the original study authors using a variety of analytical frameworks and null models.
• Sampling completeness and effort are reported as provided by the original studies and were not recalculated.
• Differences in field methodology, sampling intensity, and analytical approaches may influence reported network metrics.

This dataset is well-suited for comparative analyses, hypothesis testing related to sampling methodology and network structure, and synthesis studies across taxa and tropical regions. It is not intended to replace original datasets or interaction matrices but rather to facilitate large-scale comparisons across published seed dispersal networks.

Variable Definitions

Each row in the dataset represents a unique ecological interaction network extracted from a published study. Data that is listed as n/a is not applicable, as those data were not calculated by the original authors and presented in their manuscript. Variables are defined as follows:

Identifiers

• Comp_ID
  Unique study identifier. Studies are labeled sequentially starting from 1. Each published paper has a single Comp_ID.
• Net_ID
  Unique network identifier within a study. Networks are labeled sequentially starting from 1. Studies reporting multiple networks have multiple Net_ID values associated with the same Comp_ID.
• Site_ID
  Unique site or comparison identifier within a study (e.g., “old growth forest”, “low elevation”). Each site has a unique identifier analogous to Comp_ID.

---

Bibliographic and Geographic Information

• Citation
  Full citation of the published study from which network data were extracted.
• Country
  Country or countries where the study was conducted.
• Zone
  Ecological or biogeographic zone of the study system (e.g., tropical, temperate, riparian, montane).

---

Taxonomic Information

• Taxa
  Primary disperser taxonomic group studied (e.g., Birds, Mammals, Bats, Insects). Bats are treated as a separate category from other mammals.
• Taxa_short
  Abbreviated taxonomic category (e.g., Bi = Birds).

---

Field Methods

• Field_Methods
  Primary data collection approach used in the study, categorized as zoocentric methods, phytocentric methods, or a combination of both.
• Methods_short
  Abbreviated method classification (zoo = zoocentric, phyto = phytocentric, mix = combination of methods).
• Number_of_methods
  Total number of distinct field methods used to collect interaction data (e.g., camera trapping + mist-netting = 2).

---

Network Structure Metrics

• Net_rich
  Total network richness, including both plant and animal species.
• Plant_rich
  Number of plant species in the network.
• Disp_Rich
  Disperser (animal) species richness in the network.
• Sampling_completeness
  Sampling completeness as reported by the original study authors. This variable is included for reference and was not used in statistical analyses due to variation in calculation methods.
• Modularity_q
  Network modularity (Q) as reported in the original study.
• WNODF
  Weighted nestedness metric reported for the network.
• NODF
  Unweighted nestedness metric reported for the network.
• Connectance
  Proportion of realized interactions relative to all possible interactions in the network.
• Specialization_H2
  Network-level specialization (H2′) as reported in the original study.

---

Interaction Metrics

• Int_Freq
  Total frequency of interactions reported for the network, also referred to as edge density in some studies.
• Interaction_Diversity
  Diversity of interactions within the network as reported by the original study.
• Interaction_Evenness
  Evenness of interaction frequencies within the network as reported by the original study.
• Niche_Overlap
  Measure of niche overlap within the network as reported in the original study.
• Robustness
  Network robustness metric as reported in the original study.

Code / Software

All statistical analyses and figures associated with this dataset were conducted using R (version ≥ 4.0).

The dataset itself does not require any custom code to be interpreted or reused. Network metrics reported here were extracted directly from published studies and standardized as described in this README.

Analyses included correlation tests, nonparametric tests, and generalized linear modeling to examine relationships between network structure, network size, and field sampling methodology.

Code used for exploratory and statistical analyses is not required to reproduce the dataset and is therefore not included in this repository.