Data from: Hotspots on cold mountains: Hot flowers as pollinator refuges in mountain ecosystems

Name: Hotspots on cold mountains: Hot flowers as pollinator refuges in mountain ecosystems
Creator: Joshua Coates

Coates, Joshua 1

Published Mar 27, 2026 on Dryad. https://doi.org/10.5061/dryad.bvq83bkp6

Data files

Mar 27, 2026 version files 355.04 KB

Flower_data.csv

168.66 KB
Interaction_data.csv

174.65 KB
README.md

11.73 KB

Abstract

This dataset supports a study of floral and pollinator warming in alpine–subalpine plant communities in Kosciuszko National Park, New South Wales, Australia. The aims were to quantify variation in floral warming, and evaluate whether floral warming predicts pollinator interaction frequency. We collected infrared thermographs of flowers and plant–pollinator interactions during repeated 30-min surveys in 50 × 50 m plots. Data files provide:

IR thermograph data, with summary data from each IR image ROI, paired with ambient conditions
Plot-survey interaction counts by plant species (and pollinator order) with survey-level environmental variables. Coverage: 375 images yielding 898 flowers (43 species, 18 families) and 125 on-flower insects; 107 surveys with 1,397 interactions (24 plant species represented). Data structure, variable definitions and units are documented in the README.

Dataset DOI: 10.5061/dryad.bvq83bkp6

Corresponding author: Joshua Coates, coates.jc@hotmail.com

Description of the data and file structure

Methods

For full methods, see associated paper.

Thermal imaging

Infrared images were captured with a FLIR T420. Emissivity was set to 0.98. Images were taken from the sun‑facing side under low wind, paired with a visible image for identification. Regions of interest (ROIs) were manually defined around whole flowers and on‑flower insects.

Temperatures for each flower were processed using ThermImageJ. For each ROI we provide summary statistics (mean, SD, max, 5th and 95th percentiles) in °C.

Air temperature, relative humidity, wind speed and UV Index were measured and synchronised to each image.

Species‑level traits (flower size, colour, and shape) were compiled from Australian botanical resources; size values are mid‑points of reported ranges; colour and shape are harmonised categorical classes.

Interaction surveys

Plant–pollinator interactions were recorded during repeated 30‑min surveys within 50 × 50 m plots across three sites. For each interaction, the plant species and insect order were noted. Survey‑level environmental covariates were recorded at the start and end of each survey and averaged.

Files and variables

Thermal imagery data: 375 images, 889 flowers, 122 on‑flower insects representing 43 plant species across 18 families.
Interaction surveys represented: 104 unique surveys, 1783 interaction records involving 44 plant species.

File inventory

Flower_data.csv — Per‑ROI temperature summary values, matched environmental variables, with plant taxa information and floral traits. One row per ROI polygon (flower or pollinator) per image.
Interaction_data.csv — One row per plant–pollinator interaction event recorded during a survey, with survey‑level environmental covariates.

Variable dictionary

`Flower data.csv`

Column	Description	Type	Unit	Examples	Notes
ROI Type	Region-of-interest type (ROI) within the image	string		Flower; Pollinator
Image ID	Unique image identifier	integer		e.g., 151
Flower ID	Unique flower identifier (decimal indicates multiple flowers in same image)	number		e.g., 151.1
n pixels	Number of thermal pixels in the ROI polygon	integer		e.g., 161
mean_value	Mean temperature within ROI	number	°C
sd_value	Standard deviation of temperature within ROI	number	°C
max_value	Maximum pixel temperature within ROI	number	°C
perc95_value	95th percentile pixel temperature within ROI	number	°C
perc05_value	5th percentile pixel temperature within ROI	number	°C
Date	Image date	string	DD/MM/YYYY HH:MM	e.g., 16/12/2021 00:00
Time	Time of imaging (AEST)	number	0–1 (fraction of 24 h)	0.5 ≈ 12:00	To convert to clock time: `Time*24` hours.
wind	Wind speed	number	km h⁻¹
temperature	Ambient air temperature	number	°C
humidity	Relative humidity	number	%
uvIndex	UV Index	number	unitless index	0–11+ (here 0–8)
Flower species	Plant species name associated with IR Image	string		e.g., Bossiaea foliosa
family	Plant family	string		e.g., Fabaceae
flower_size_mean_mm	Mean flower size (midpoint of reported size range)	number	mm
flower_colour_category	Standardised flower colour category	string	—	blue; pink; purple; white; yellow
flower_shape_category	Standardised flower morphology category	string	—	bell-shaped; cup-shaped; globular head; radiate head; star-like; tubular; umbel/cluster; pea; spider-like; other

`Interaction data.csv`

Column	Description	Type	Unit/Encoding	Allowed values / examples
Plant species	Plant species receiving the interaction	string		e.g., Taraxacum officinale
Plant Family	Plant family of plant species	string		e.g., Asteraceae
Insect Order	Taxonomic order of visiting insect	string		e.g., Diptera; Hymenoptera; Coleoptera
Wind	Wind speed during survey	number	km h⁻¹
Temperature	Ambient air temperature during survey	number	°C
Humidity	Relative humidity during survey	number	%
Elevation	Elevation of the site	string	m a.s.l.	e.g., 1600m; 1800m; 2000m
Date	Survey date-time in Excel serial format	number	Excel serial date (days since 1899‑12‑30)	e.g., 44511.04167 ≈ 2021‑11‑11 01:00
Survey ID	Unique identifier for 30‑min plot survey	integer	—	e.g., 73
Time	Survey start time as fraction of day	number	0–1 (fraction of 24 h)	0.5 ≈ 12:00

Code/software

Temperature extraction: ImageJ, ThermImageJ (ImageJ plugin).

Statistical analysis: RStudio with common packages (e.g., tidyverse, glmmTMB, mgcv, emmeans).

How these data map to the analyses

Species‑level warming

Aggregate ROI records by Flower species to test relationship between floral species and ΔTmax

Environmental drivers

Model ROI‑level temperature metrics against per‑image environmental variables (uvIndex, temperature, humidity, wind) with appropriate random effects for plant species/image as described in paper.

Floral traits

Join ROI records to trait columns to test relationship between traits and ΔTmax

Interaction rates

Use Interaction data.csv at the survey level and join species‑level ΔTmax classes to plant species to evaluate temperature‑dependent interaction frequency patterns.