Data and code from: Non-consumptive effects of odonate predators on pollinators of water lilies
Data files
Sep 18, 2026 version files 60.17 KB
-
Code_All.R
26.61 KB
-
dataset_field_experiment.csv
30.58 KB
-
README.md
2.98 KB
Abstract
Plant-pollinator interactions are shaped not only by environmental conditions but also by biotic interactions such as predation. Predators influence prey through direct consumption and non-consumptive effects (NCEs) arising from perceived predation risk, altering prey traits such as foraging behaviour. Emergent freshwater macrophytes such as the yellow water lily (Nuphar lutea) create a shared interface where predators and pollinators co-occur. Yet, NCEs of predation risk on pollinators and their consequences for plant-pollinator interactions remain severely understudied, especially at aquatic-terrestrial transition zones. Therefore, we experimentally tested how artificial dragonflies, as standardised visual predation cues, shape the foraging behaviour of pollinators on N. lutea and assessed whether these behaviours were related to natural variation in odonate abundance. We quantified visitation rates and visit durations under artificial dragonfly, neutral object, and control treatments, and related visit duration to natural odonate abundance. Our results showed that artificial dragonflies reduced visitation rates by up to 50 % and visit durations by up to 18 seconds per visit, across pollinator taxa. Visit duration showed a negative association with natural odonate abundance, and structural equation models revealed that pollinator behaviour was indirectly associated with environmental conditions via their effects on odonate abundance. Consequently, our findings demonstrate that predator-mediated NCEs play a substantial role in shaping pollinator foraging on N. lutea, highlighting that plant-pollinator interactions at freshwater-land ecotones may depend on dynamic predation risks and underlying environmental drivers. These insights are particularly important for understanding macrophyte-insect interactions because communities at aquatic-terrestrial interfaces are especially vulnerable to shifts in environmental conditions.
Dryad DOI: https://doi.org/10.5061/dryad.1c59zw4bt
Ole Rieken, Rebecca Oester
This README file contains information on the R scripts and variables in the datasets used for the analyses.
Data file
dataset_field_experiment
Main dataset with information on each pollinator visit with the following variables:
- Flower_ID: Unique identifier for individual flowers within each observational block
- Date: Date of observation
- Time: Time of observation
- Temp: Ambient air temperature at the time of observation (in °C)
- Light_Cond: Light conditions at the flower (shade, half-shade, sunny)
- Weather: General weather conditions (cloudy, clear, rain)
- Wind: Presence of noticeable wind during observation (yes/no)
- Odonate_Abundance: Number of odonates present during a pollinator visit (individuals/m^3)
- Pollinator_Group: Pollinator group (order level: Hymenoptera, Diptera, Coleoptera...)
- Visit_Duration(s): Visit duration (in seconds)
- Treatment: Experimental treatment applied to the flower (Artificial dragonfly, Control, Neutral object)
- Obs_Block: Observational block (one hour each)
- Area: Study area identifier indicating spatial subdivision of the site (A or B)
- Notes: Additional notes, including species or family identification and noteworthy observations
R-script
Run in R (version 4.5.1) on Linux.
Code_All.R
requires dataset_field_experiment.csv
This code file contains all code used for our analyses and is organized into four main sections:
1) We used generalized linear mixed models (GLMMs) to test how treatments affected visitation rates of Diptera and Hymenoptera, with observational blocks as random effects. Treatment differences were assessed using estimated marginal means and visualized. Differences between pollinator groups were tested using treatment × group interactions.
2) We used linear mixed models to test how experimental treatments and odonate abundance affected visit duration, with observational blocks as random effects. We compared models using AIC and Likelihood Ratio Tests, including comparisons between pollinator groups. Treatment differences were assessed using estimated marginal means and visualized in boxplots.
3) We examined the relationship between odonate abundance and pollinator visit duration using Pearson correlations for each pollinator group. We also visualised treatment effects and interactions with odonate abundance using predicted values from the LMMs.
4) We used structural equation models (SEMs) to examine indirect effects of wind and weather on pollinator visit duration through odonate abundance, modelling Diptera and Hymenoptera separately. Visit duration was log10-transformed, and model fit was assessed using multiple fit indices and theoretical expectations.
To investigate possible effects of artificial dragonflies on pollinator behaviour (visitation rate, visit duration), we performed a field experiment using a randomised block design. We observed a total of 22 one-hour blocks over a span of 15 different days between June and August. This corresponds to an observational period of 1320 minutes. We selected nine yellow water lily flowers per block and observed visits by pollinating insects. Visit duration was measured using a stopwatch. We assigned the treatments randomly to individual flowers, resulting in three flowers per treatment (artificial dragonfly, neutral object, control). To examine a potential relationship between natural odonate abundance and visit duration, we recorded the number of odonates active in the immediate vicinity of a flower during each pollinator visit (individuals/m3). We minimised environmental differences between blocks as much as possible to ensure the most independent distribution and observation of the treatments. Wind and weather were recorded as binary environmental variables with “wind” as ~2 m/s or higher or “no wind” when calm. We recorded weather as either overcast or clear, as these were the only conditions where pollinators of N. lutea were effectively active.