Data from: Pollinator assemblage composition predicts trait divergence in a pollination-generalized plant
Data files
Mar 23, 2026 version files 8.45 MB
-
Flower_Pollinator_Divergence.zip
8.44 MB
-
README.md
8.02 KB
Abstract
The causal role of pollinators in driving the divergence of plant traits is a fundamental tenet of angiosperm evolution, providing hallmark examples of natural selection. However, it remains unclear how geographic variation in pollinator assemblages relates to the divergence of pollination traits in pollination-generalized plants. We characterized pollinator assemblages that interacted with Viscaria vulgaris in southern Sweden, and evaluated, through statistical dimension reduction, whether pollination traits were associated with an inferred main axis of geographic variation in pollinator assemblages. We documented a functionally broad range of pollinators that visited V. vulgaris. Although the most frequent pollinator functional groups were present in most populations, their relative contribution to flower visitation varied across the study area, establishing a geographic mosaic of local pollinator assemblages. We demonstrate that the geographic variation of local pollinator assemblages can predict the divergence of pollination traits in V. vulgaris. The findings of this geographic comparative study are consistent with the hypothesis that geographic variation in pollinator assemblages drives the divergence of pollination traits in pollination-generalized plants. Thus, generalized plant-pollinator interactions do not preclude the divergence of pollination traits, which may maximize the collective contribution of local pollinator assemblages rather than that of a principal pollinator.
Dataset DOI: 10.5061/dryad.xwdbrv1tf
Description of the data and file structure
Data were collected from 27 populations of Viscaria vulgaris in southern Sweden between 2021 and 2024. At the inflorescence and flower level, we measured traits related to pollinator attraction (inflorescence length, flower number, corolla diameter) and pollen transfer (tube width and length, nectary–stigma distance) on 3,878 inflorescences from 1,797 individuals. Concurrently, repeated 10-minute censuses recorded the number and identity of flower visitors, which were classified into functional groups based on morphology and foraging behaviour. These data captured spatial, temporal, and developmental variation in both plant traits and pollinator activity, forming the basis for multivariate statistical analyses linking divergence in pollination traits to variation in local pollinator assemblages.
Files and variables
The repository contains the zipped folder Flower_Pollinator_Divergence.zip which is organized into three main folders: code, data and R_Functions.
1. The code folder. This folder contains R scripts used to analyze the data, including:
- Flower_Traits_Variance_Partitioning.Rmd: Performs variance partitioning of pollination traits using multivariate generalized linear mixed-effects models (GLMMs) in
Hmsc. This script also contains code for estimating the proportional divergence (dP) of pollination traits. - Pollinators_Variance_Partitioning.Rmd: Performs variance partitioning of pollinator assemblages using multivariate generalized linear mixed-effects models (GLMMs) in
Hmsc. - RRR_Abundance_Population.Rmd: Conducts reduced-rank regression (RRR) to relate divergence in pollination traits to the composition of local pollinator assemblages in
Hmsc. - RRR_Abundance_Population_Graphs.Rmd: Contains code to reproduce the graphs of the reduced-rank regression (RRR) analysis. It quantifies the contribution of pollinator functional groups to the 'pollinator assemblage axis' and includes the graphical outputs summarizing the relationships of pollinator assemblage composition (the 'pollinator assemblage axis') and pollination traits as obtained from the RRR analysis.
- RRR_Predictive_Explanatory_Power.Rmd: This code evaluates the predictive and explanatory performance of the reduced-rank regression (RRR) models.
Running these scripts reproduces the main statistical analyses described in the associated manuscript.
2. The data folder. This folder contains raw and processed datasets used in the main analyses. The measurements are in mm. Files include:
- Data_Pollination_Traits_Master.rds: Master file containing measurements of pollination traits.
- Population: Identifier for the population
- Year: The year in which the data were collected
- Date: The specific observation date
- Plant_ID: Unique identifier for each individual plant
- Stem_Height: Inflorescence Length
- Stickidium: Glandular Region Length
- Total_Flowers: Total Number of Flowers Produced
- Corolla_Diameter: Width of the Flower across the Corolla
- Tube_Width: Width of the Floral Tube
- Tube_Length: Length of the Floral Tube
- NSD: Nectary-Stigma Distance
- log_Stem_Height: log-transformed Stem Height
- log_Stickidium: log-transformed Stickidium
- log_Total_Flowers: log-transformed Flower Number
- log_Corolla_Diameter: log-transformed Corolla Diameter
- log_Tube_Width: log-transformed Tube Width
- log_Tube_Length: log-transformed Tube Length
- log_NSD: log-transformed Nectary-Stigma Distance
- Buds_Traits: Number of Flower Buds Produced
- Buds_Eaten_Traits: Number of Buds that were Eaten/Damaged
- Capsules_Traits: Number of Mature Capsules Produced
- Data_Pollinators_Master.rds: Master file containing pollinator census data.
- Unique_Observation_ID: Unique identifier for each census
- Population: Identifier for the population
- Year: Year in which the census took place
- Date: Calendar date of the census
- Time_Start: Start time of the census
- Each of the following variables represents the number of visits recorded for a given pollinator group during each census:
- Bombus_Queen: Visits by queen bumblebees
- Bombus_terrestris_Worker: Visits by worker individuals of Bombus terrestris
- Bombus_Worker: Visits by worker bumblebees (other Bombus species)
- Bombyliidae: Visits by bee flies (family Bombyliidae)
- Coleoptera: Visits by beetles
- Diurnal_Lepidoptera: Visits by day-active butterflies
- Honeybee: Visits by honeybees (Apis mellifera)
- Large_Solitary_Bee: Visits by large-bodied solitary bees
- Noctuidae: Visits by moths of the family Noctuidae
- Other: Visits by pollinators not classified into the listed groups.
- Small_Solitary_Bee: Visits by small-bodied solitary bees
- Sphingidae: Visits by moths (family Sphingidae)
- Syrphidae: Visits by hoverflies (family Syrphidae)
- Total_Visits: Total number of pollinator visits recorded across all groups during each census
- Sky: Description of sky conditions during observation (e.g., clear, partly cloudy, overcast).
- Temperature: Temperature during the observation period (e.g., categorical scale)
- Wind: Wind conditions during observation (e.g., categorical scale)
- RRR_Model_Abundance.rds: Output from reduced-rank regression analysis relating pollinator assemblage composition to pollination trait divergence.
- RRR_Pollinator_Assemblage_Axis.rds: This file contains the two main reduced axes derived from the reduced-rank regression (RRR) analysis. It stores the population-level scores for the first and second 'pollinator assemblage axis'. These axes represent the gradients of geographic variation in pollinator assemblage composition that best explain divergence in pollination traits. The values correspond to the reduced composite variables constructed as linear combinations of pollinator functional groups during the RRR analysis.
3. The R_Functions folder. This folder contains a R function required to run specific components of the RRR analyses. constructGradientRRR.R defines the R function used to construct predictor gradients along the reduced-rank regression (RRR) axes.
4. Missing values. Missing or unavailable measurements are indicated as NA. Blank cells are not used to represent missing values to ensure compatibility with R functions.
5. Additional files generated automatically during project setup: .git is a hidden folder created by Git to track version history and changes for this project. .Rproj is a RStudio project file that stores project-specific settings, working directory, and environment preferences.
Code/software
All data can be viewed and analyzed using the free and open-source software R. Analyses and data handling largely rely on packages for data import and manipulation (readr, dplyr, tibble, tidyr, reshape2, stringr), visualization (ggplot2, ggpubr), statistical modeling (Hmsc), and general data organization. These packages allow users to reproduce the data cleaning, visualization, and multivariate analyses conducted in the study.