Pollinator efficiency, rather than bee decline, explains an shift to hummingbird pollination in tropical montane forests
Data files
May 06, 2026 version files 51.88 KB
-
Data.Availability(Dryad).zip
38.96 KB
-
R-codes(Dryad).zip
6.11 KB
-
README.md
6.81 KB
Abstract
In animal-pollinated plants, adaptive floral divergence to distinct pollination environments can lead to pollination shifts, promoting the evolution of new pollination syndromes that confer reproductive isolation. A longstanding but untested hypothesis proposes that reduced bee visitation in tropical montane cloud forests has repeatedly driven the evolution of hummingbird-pollinated plant species in diverse lineages. Here, we test whether recently diverged bee and hummingbird-pollinated syndromes in two sister species are locally adapted to their native pollination environments, and whether this adaptation reflects declining bee activity at higher elevations. Alternatively, we ask whether higher pollen transfer efficiency can drive adaptation to hummingbirds regardless of bee availability. We measured visitation and per-visit pollen deposition to calculate pollinator effectiveness as a quantitative measure of pollinator performance, and conducted reciprocal translocations of Costus kuntzei, with ancestral bee pollination, and Costus wilsonii, with derived hummingbird pollination, across an elevational gradient in Costa Rica, including sites within and outside each species’ range as well as at their elevational boundary. In their respective ranges, the species are specialized on their respective pollinator functional groups, bees or hummingbirds. However, C. wilsonii received more pollen grains per visit despite C. kuntzei experiencing higher visitation rates. Consequently, pollinator effectiveness was greater for C. wilsonii. In reciprocal translocations, C. kuntzei showed uniform bee visitation and effectiveness across habitats and elevations, whereas hummingbird visitation and effectiveness increased with elevation for C. wilsonii. Our results show that pollination systems are adapted to their native pollination environments, and that floral adaptation to hummingbird pollination is likely driven by higher hummingbird visitation in montane environments combined with their higher per-visit pollen deposition efficiency, rather than by a decline in bee visitation with elevation as previously assumed. This study provides the first thorough experimental test of the drivers of evolutionary transitions to hummingbird pollination in the American tropics.
Dataset DOI: 10.5061/dryad.931zcrk0r
Description of the data and file structure
The repository is organized into two main folders corresponding to the data (Data.availability (Dryad)) and the R codes (R-codes (Dryad)) used for the analyses presented in the paper.
Files and variables
File: Data.Availability(Dryad).zip
Description: This folder (Data.Availability(Dryad).zip) contains our data and all the necessary files to replicate the analyses.
(1) wild.xlsx: This dataset contains wild-Costus flower pollinator visitation observations collected across multiple years, sites, and species categories. Each row corresponds to one observation record and includes date, year, site, locality class, observation ID, observation duration, log10-transformed duration, species code, and total recorded visits.
File name: wild.xlsx
Description: Pollinator visitation observations from wild flowers across sites and years.
Number of rows: 143
Number of columns: 9
Worksheet: wild
Variables:
Date: sampling date; currently stored using mixed numeric encodings
yr: sampling year
Site: site code (LA, LC, LG). LA is Las Alturas (high elevation site); LC is Las Cruces (mid elevation site); LG is La Gamba (low elevation site).
loc: locality category (high, low)
ID: observation or flower identifier
Duration: observation duration in hours
log: log10-transformed observation duration
species: species code (K, W). K is Costus kuntzei and W is Costus wilsonii.
visits: number of pollinator visits recorded during the observation period
(2) pervisit.xlsx: This dataset contains pollinator visitation and pollen deposition observations from individual stigmas of wild Costus flowers collected across multiple sites and species categories. Each row corresponds to one stigma-level observation record and includes date, observation duration, log10-transformed duration, site of origin, locality class, species code, stigma ID, number of pollen grains deposited, number of pollinator visits, and pollen deposition efficiency.
File name: pervisit.xlsx
Description: Pollinator visitation and pollen deposition data from individual stigma observations across sites and years.
Number of rows: 97
Number of columns: 10
Worksheet: pervisit
Variables:
date: sampling date, stored as numeric values
duration: observation duration in hours
log_duration: log10-transformed observation duration
site_origin: site code for the observation or source flower (LA, LC, LG). LA is Las Alturas (high elevation site); LC is Las Cruces (mid elevation site); LG is La Gamba (low elevation site).
loc: locality category (high, low)
species: species code (K, W). K is Costus kuntzei and W is Costus wilsonii.
id_stigma: stigma or observation identifier
no_pollen_grains: number of pollen grains recorded on the stigma
no_visits: number of pollinator visits recorded during the observation period
efficiency: pollen deposition efficiency, calculated as pollen grains per visit
(3) HummingBee.xlsx: This dataset contains pollinator visitation observations from experimental floral arrays of Costus collected across multiple years and sites. Each row corresponds to one array-level observation record and includes sampling year, site, array identifier, observation duration in hours, log10-transformed duration, and the total recorded visits to species K and species W.
File name: HummingBee.xlsx
Description: Pollinator visitation data from experimental floral arrays of Costus species across sites and years.
Number of rows: 102
Number of columns: 7
Worksheet: HummingBee.PJuarez
Variables:
yr: sampling year
site: site code (LA, LC, LG). LA is Las Alturas (high elevation site); LC is Las Cruces (mid elevation site); LG is La Gamba (low elevation site).
array_ID: identifier for each floral array observation
array_hours: duration of array observation in hours
array_log: log10-transformed observation duration
K: number of bee visits recorded to species Costus kuntzei during the observation period
W: number of hummingbird visits recorded to species Costus wilsonii during the observation period
File: R-codes(Dryad).zip
Description: Folder containing the R scripts used to analyze pollinator visitation, per-visit pollen deposition efficiency, and overall pollinator effectiveness in Costus kuntzei and C. wilsonii, including comparisons across localities and reciprocal translocation sites. The scripts read the archived datasets wild.xlsx, pervisit.xlsx, and HummingBee.xlsx, fit generalized linear mixed models, estimate marginal means and pairwise comparisons, generate figures, and in calculate pollinator effectiveness using bootstrap resampling of per-visit efficiency values.
Files in this folder:
(1) COMPARING POLLINATOR EFFECTIVENESS COMPONENTS IN C.KUNTZEI & C.WILSONII.R: R script for comparing pollinator visitation rate, per-visit efficiency, and overall pollinator effectiveness between C. kuntzei and C. wilsonii across locality classes using wild.xlsx and pervisit.xlsx. The script fits GLMMs, computes estimated marginal means and Sidak comparisons, generates plots, and bootstraps pollinator effectiveness as visitation × efficiency.
(2) LOCAL FLORAL ADAPTATION_C.KUNTZEI & C.WILSONII TO POLLINATOR VISITATION AND EFFECTIVENESS.R: R script for analyzing local floral adaptation in a reciprocal translocation experiment using HummingBee.xlsx and pervisit.xlsx. The script reshapes visitation data from floral array-level counts, models pollinator visitation and pollinator effectiveness across sites/elevations, computes estimated marginal means and pairwise comparisons, generates plots, and bootstraps pollinator effectiveness from per-visit efficiency values.
Code/software
All data can be viewed and analyzed using the free and open-source software R statistical environment version 4.4.1. Analyses and data handling rely on packages for data import and manipulation (readxl, tidyverse), visualization (ggplot2, ggeffects), and statistical analyses (glmmTMB, emmeans). These packages allow users to reproduce the data visualization and analyses conducted in our study.
Access information
Other publicly accessible locations of the data:
- None
Data was derived from the following sources:
- Field observations and experiments.