Pollinator interactions of native and introduced plants in smallholder tropical orchards across a gradient of anthropogenic landscapes
Data files
Aug 13, 2025 version files 221.59 KB
-
Dataset.zip
108.61 KB
-
NW_level.zip
8.31 KB
-
README.md
5.30 KB
-
SP_and_NW_level-Pollination_network_each_site.R
36.98 KB
-
SP_level.zip
62.39 KB
Abstract
Anthropogenic habitats commonly favor introduced species. However, there is little information regarding the responses of pollinators to native and introduced plants across anthropogenic landscapes. The main goal of the study is to investigate pollinator interactions of native and introduced plants across smallholder tropical orchards in Thailand, as these plantations are essential to local food security.
We compiled data from four recent studies on plant-pollinator networks in 65 smallholder tropical orchards across Thailand. Using network analyses, we compared species degree (i.e., the number of pollinator species visiting a plant) and specialization (d′) between native and introduced plant species, with the latter further categorized into non-naturalized, naturalized, and invasive groups. Analyses were conducted using linear mixed-effects models, treating individual plant species as the unit of analysis and incorporating orchard identity nested within study as a random effect. Additionally, we examined whether variation in surrounding anthropogenic landscapes (i.e., the proportion of agricultural and urbanized areas) and the proportional representation of introduced plant species influenced the structure of plant–pollinator networks.
Overall, specialization (d′) of the native plants was significantly greater than that of the introduced plants, while pollinators exhibited no preference towards naturalized or invasive plants over non-naturalized ones. We found a decrease in species degree of native plants as well as connectance and linkage density with increasing proportion of anthropogenic areas, likely due to limited floral resources and nesting availability. However, we also found an increase in network specialization (H2′) with a higher proportion of urban area.
Understanding the impact of land-use changes with the presence of introduced plant species is important for predicting community stability of plant-pollinator interactions as well as for habitat management strategies. Surrounding natural habitats and important food plants should be preserved and promoted to maintain pollinator interactions in human-dominated areas.
Dataset DOI: 10.5061/dryad.0cfxpnwdx
Description of the data and file structure
This project analyzes how plant-pollinator interaction networks vary across different plant categories and landscapes in tropical orchards. We focus on both species-level and network-level metrics to assess the ecological roles of native vs. introduced plants, human food plants, and the impacts of land-use change.
Files and variables
File: Dataset.zip
Description: Compiled CSV datasets of plant–pollinator interactions across 96 orchard sites. Used as input for metric calculations.
File: SP_and_NW_level-Pollination_network_each_site.R
Description: R script to calculate species-level and network-level pollination metrics per site.
File: SP_level.zip
Description: Contains all scripts and data for species-level analyses, including group comparisons and land-use effects.
File: NW_level.zip
Description: Contains all scripts and data for network-level analyses, focusing on land-use impacts on connectance, linkage density, and H2′.
Code/software
SP and NW level-Pollination network each site.R
This script calculates both species-level and network-level pollination metrics for each orchard site using binary plant-pollinator interaction matrices.
Species-Level Metrics
- Computed using
specieslevel()from thebipartitepackage.
Network-Level Metrics
Computed for each site using networklevel():
- Connectance
- Nestedness
- Linkage density
- H2′ (network-level specialization)
Workflow Summary
- Loads 1–96 site-specific CSV files from
Dataset.zip, each representing a unique plant-pollinator interaction matrix. - For each site:
- Reads the matrix.
- Calculates species- and network-level metrics.
SP level.zip
LMM-comparison plant groups.R
This script performs comparative statistical analysis of species-level pollination network metrics across five plant group contrasts using linear mixed-effects models (LMMs).
Plant Group Comparisons
- Native vs. Introduced
- Non-naturalized vs. Naturalized
- Naturalized vs. Invasive
- Human-food vs. Non-human-food
- Native human-food vs. Introduced human-food
Metrics Analyzed
- Species Degree (log-transformed)
- Specialization (d′) (square-root transformed)
Workflow Summary
- Loads preprocessed datasets from
LMM-plant groupsfolder. - Transforms metrics to improve normality based on Shapiro tests.
- Fits LMMs for each comparison with
Study_ID / Orchard_IDas nested random effects. - Extracts model summaries and calculates:
- Marginal R² (variance explained by fixed effects)
- Conditional R² (variance explained by fixed + random effects)
- Applies Benjamini–Hochberg correction within each metric for multiple testing control.
LMM-SP level-Land use.R
This script investigates the effects of land-use change on species-level pollination network metrics — species degree and specialization (d′) — for native and introduced plants, using linear mixed-effects models (LMMs).
Key Objectives
- Assess how increasing proportions of agricultural (AG) and urbanized (UR) landscapes around orchards influence plant species’ interaction metrics.
- Evaluate native and introduced plants separately to detect differential land-use responses.
Dataset
- Input:
LMM-Dataset-Native-Intro-Land use-Final.csv - Transformed metrics:
log_degree= log(species_degree)sqrt_dprime= sqrt(d′)
- Nested random effects:
Orchard_IDwithinStudy_IDto account for spatial and study-level variation.
Modeling Workflow
- Separate models for native and introduced plant groups:
- Full model: AG + UR
- Reduced models: AG-only, UR-only, and null models
- Model selection via likelihood-ratio tests using
anova()fromlme4. - R² values computed using
r.squaredGLMM()from theMuMInpackage.
NW level.zip
LMM-NW level-Landuse.R
This script analyzes how landscape composition (agricultural and urbanized areas) and the proportion of introduced plants influence network-level pollination metrics in smallholder tropical orchards using linear mixed-effects models (LMMs).
Objective
To examine the effects of:
% Agricultural area (AG)% Urbanized area (UR)% Introduced plant species (INTRO)- And their interactions
on three key network-level metrics:
- Connectance
- Linkage Density (log-transformed)
- Specialization (H2′)
Dataset
- Input:
Dataset-NW level-Land use-Final.csv - Random effect:
Study_IDto account for variation across studies.
Modeling Workflow
- Stepwise model selection via likelihood-ratio tests (LRT) using
anova()to compare nested LMMs. - For each metric, models start from full interaction models (AG × INTRO, UR × INTRO), with stepwise simplification.
- Model assumptions assessed using
check_model(). - Goodness-of-fit measured via AIC, logLik, R² marginal (fixed effects), and R² conditional (fixed + random effects) using
MuMIn::r.squaredGLMM().