Leaf and canopy temperatures
Observed Tleaf and leaf-to-air temperature difference, ΔT, across species and between provenances, while under common microclimatic conditions, were determined using two different methods: first by tracking individual Tleaf with point measurements taken over 10 days using an infrared thermometer, and subsequently by determining whole canopy temperature (Tcan) using UAV- based thermal imagery. Individual leaf temperatures were recorded using a dual laser infrared thermometer (MS6530, Mastech, Pittsburgh, Pennsylvania, US), with a spectral response of 8-14 µm, a distance to spot size ratio of 12:1, and an assumed emissivity at 0.95. Measurement campaigns for Tleaf took place between the times of 10:45 and 14:45 during cloud-free periods with low windspeed. Measurements were standardised across trees by taking measurements on 10 sun-exposed leaves on branches that were perpendicular to the solar angle to ensure maximum incoming radiation. This was repeated over multiple days and in a randomised order. The total number of leaves measured on each plant ranged from 30 to 50 (n = 1990). The difference between leaf and air temperatures (ΔT) for each plant was calculated by subtracting 10-min average air temperature.
For Tcan, thermal images of whole tree canopies were taken using a drone (DJI Mavic 2 Enterprise Advanced, DJI Technology Co. Ltd, Shenzhen, China) fitted with a FLIR M2ED thermal camera (Teledyne Flir, Wilsonville, OR, USA) with a thermal spectral band ranging from 8 to 14 µm, an uncooled Vox microbolometer thermal sensor (with automated calibration), sensor resolution of 640 x 512 pixels, approx. 9 mm lens with a 57° horizontal field of view, and 30 Hz frame rate. The flight was completed on 11th December 2023 at 12:07 pm, with a 15 min flight time. Both thermal (IR) and true colour (RGB) images were taken simultaneously and flown with 50% overlap at 1.3 m s-1. With a flight altitude of 27 m above the ground the pixel resolution was 4.2 cm pixel−1. RGB images were captured at the same times and locations as the thermal images, with an image size of 5472 x 3648 pixels, 35 mm format equivalent lens length of 24 mm, and a horizontal field of view of 84°. An RGB orthophoto mosaic of the trial site with an average ground sample distance of 2.18cm was created using Pix4Dmapper (Pix4D S.A., Prilly, Switzerland).
Thermal images were converted to TIFF, with radiometric data converted to temperatures using the mean air temperature (29 °C) and relative humidity (38 %) measured on the crane weather station during the flight, and an object distance of 27 m minus the target tree height, resulting in object distances ranging from 18.00 m to 26.34 m (mean 22.54). Images were analysed using imageJ, with target tree canopies determined manually using polygons, and background temperatures of the ground excluded using threshold filtering. For each tree, calculations of the mean, standard deviation, median, minimum, maximum, and skewness of canopy temperatures were recorded. We used the mean of each individual canopy for analysis of Tcan.