Data and code from: Trapped honey bees reduce floral visitation on milkweed flowers
Data files
Apr 24, 2026 version files 27.66 KB
-
README.md
5.32 KB
-
trapped_bee_data_and_analysis.zip
22.35 KB
Abstract
Trapped arthropods have been shown to benefit plants in several ways, but few studies have examined the potential costs of arthropod entanglement. Milkweeds (Asclepias spp.) have an unusual pollination system that requires relatively large packets of pollen (pollinia) to become attached to the appendages of insect pollinators and be pulled through a narrow opening in the flower. Honey bees (Apis mellifera) commonly become trapped and die with their legs still attached to milkweed flowers. In this study, we conducted a field experiment to examine how dead trapped honey bees affect floral visitation. We expected that the presence of a dead trapped bee would reduce floral visitation via two non-mutually exclusive pathways: 1) a direct deterrent effect on floral visitors, and 2) an indirect deterrent effect mediated by an increased abundance of scavenging predators such as ants. The presence of a dead bee reduced floral visitation by 37% compared with controls, and this effect was more robust for honey bee visitors than non-Apis visitors. While ant densities were 51% higher on floral umbels with a dead bee, and ants reduced floral visitation by 30%, our path analysis indicated that the direct deterrent pathway explained 91% of the total effect, consistent with an aversion to dead conspecifics among honey bees. Our results suggest that the lethal entanglement of honey bees is likely to incur an ecological cost for milkweed flowers, although the deterrence of honey bees could also shift the pollinator communities on milkweeds with unexpected consequences.
Access this dataset on Dryad: https://doi.org/10.5061/dryad.w9ghx3g3j
These are CSV data files and RMD analysis scripts from a project that conducted a field experiment to examine how dead trapped honey bees affect floral visitation.
Description of the data and file structure
Files and variables
File: trapped_bee_data_and_analysis.zip
Compressed folder including two data files (CSV) and an analysis script in Rmarkdown.
Description of the data files
Each observation video was assessed on both the near scale and the umbel scale.
- "near.scale" refers to data collected based on floral visitation to the focal flower (carcass/control) and a radius of two flowers extending away.
- "umbel.scale" refers to floral visitation data collected based on all flowers within the umbel.
For our analysis, we focused on the umbel scale dataset as it seems more intuitive to analyze floral visitation on the scale of the entire umbel and exploratory results were nearly identical. However, we are making the near.scale dataset available for future researchers.
"Paired Plant Data 2023-2024 [near.scale] 2025-07-08.csv": paired floral visitation at the "near" scale.
- date: date in mm/dd/YYYY format
- temp.C: air temperature in degrees C. Temperatures were not recorded in 2023, and blank cells should be interpreted as "NA" values.
- trial: numerical pair id within year (replaced with unique pair.id in overall analysis)
- start.time: observation start time
- obs.dur: observation duration in mm:ss
- time.recorded: text description of observation start and end times
- trt: "Carcass" = "dead bee present" and "Control" = "control"
- plot.ID: site-specific plot ID; plots are 5 x 5 m areas defined for another project. Plot IDs were not recorded in 2023, and blank cells should be interpreted as "NA" values.
- plant.ht.cm: plant height in cm. Plant heights were not recorded in 2023, and blank cells should be interpreted as "NA" values.
- flower.count: number of flowers in each focal umbel
- visitor.count: total floral visitor count
- apis.count: count of Apis mellifera visitors
- other.visitors.count: count of non-Apis visitors
- ant.count: the number of ants visible on the umbel at each minute mark, summed across the observation
- notes: additional notes
"Paired Plant Data 2023-2024 [umbel.scale] 2025-07-08.csv": paired floral visitation at the "near" scale.
- date: date in mm/dd/YYYY format
- temp.C: air temperature in degrees C. Temperatures were not recorded in 2023, and blank cells should be interpreted as "NA" values.
- trial: numerical pair id within year (replaced with unique pair.id in overall analysis)
- start.time: observation start time
- obs.dur: observation duration in mm:ss
- time.recorded: text description of observation start and end times
- trt: "Carcass" = "dead bee present" and "Control" = "control"
- plot.ID: site-specific plot ID; plots are 5 x 5 m areas defined for another project. Plot IDs were not recorded in 2023, and blank cells should be interpreted as "NA" values.
- plant.ht.cm: plant height in cm. Plant heights were not recorded in 2023, and blank cells should be interpreted as "NA" values.
- flower.count: number of flowers in each focal umbel
- visitor.count: total floral visitor count
- apis.count: count of Apis mellifera visitors
- other.visitors.count: count of non-Apis visitors
- ant.count: the number of ants visible on the umbel at each minute mark, summed across the observation
- other.bee.count: count of non-Apis bee visitors
- spider.count: count of spider (Araneae) visitors
- wasp.count: count of wasp visitors
- tachinid.count: count of tachinid fly (Tachinidae) visitors
- bombyliidae.count: count of bee fly (Bombyliidae) visitors
- syrphid.count: count of hover fly (Syrphidae) visitors
- lepidoptera.count: count of butterfly and moth (Lepidoptera) visitors
- hemiptera.count: count of true bug (Hemiptera) visitors
- lacewing.count: count of lacewing (Neuroptera) visitors
- notes: additional notes
"Bee Carcass Milkweed Project Analysis 2026-04-23.Rmd": analysis script in Rmarkdown. Running this code will generate the complete analysis and figures. This script uses the following packages:
- library(tidyverse) for general data analysis
- library(lme4) for linear mixed modeling
- library(DHARMa) for testing overdispersion
- library(glmmTMB) for zero-inflated models
- library(lavaan) for path analysis
- library(lavaanPlot) for plotting path analysis
- library(MuMIn) for multimodel selection using model.sel()
- library(patchwork) for multipanel plotting
- library(DiagrammeRsvg) for required to export path analysis plot
- library(rsvg) for required to export path analysis plot
- library(knitr) for formatting tables with kable
- library(lmeresampler) for bootstrap resampling
Code/software
All analyses were conducted in R version 4.5.0.
R Core Team. 2025. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
