COS 290-4 - Mapping tropical tree species canopy temperatures using UAV thermal, hyperspectral, and LiDAR imaging

Abstract: Canopy and leaf temperatures are tightly linked to biological and ecological processes including plant growth, stress, phenology, and rates of productivity. However most studies are limited to single leaves, ignoring potential differences in temperatures within a crown, or are measuring entire canopies where species differences are difficult to resolve. Thermal infrared (TIR) imaging from unoccupied aerial vehicles (UAVs) can be used to map canopy and leaf temperatures for different species across large extents. The goals of our study were to (1) map tree crowns within a 4-hectare montane wet forest site, (2) quantify species-level differences between canopy temperatures and air temperatures (Tdiff) throughout the day, and (3) quantify species-level responses (NDVI, EVI, FCVI, and NIRv) to temperatures (air, canopy, Tdiff). To achieve these goals, we used GatorEye Unmanned Flying Laboratory (UAV) to conduct repeat flyovers of a four-hectare montane wet forest field site at Laupāhoehoe, Hawai’i, collecting high spatial resolution (~20cm) hyperspectral, thermal, lidar and visual data. We mapped two dominant species in the canopy as ‘ōhi‘a lehua (Metrosideros polymorpha, N=311) and koa (Acacia koa, N=103) though two other trees reach canopy or sub-canopy height including ʻōlapa (Cheirodendron trigynum, N=44) and pilo (Coprosma rhynchocarpa, N=18). For each time point, and for each tree crown, we calculated multiple vegetation indicators to be used as proxies for gross primary productivity including NDVI (normalized difference vegetative index), EVI (enhanced vegetative index), FCVI (fluorescence correction vegetation index), and NIRv (near infrared reflectance), and Tdiff. We show significant differences in crown temperatures and Tdiff by species – ‘Ōhi‘a lehua had the highest crown temperatures (17.4 C) and deviated the least from air temperature (by about 1.2 C). We found the strongest positive relationships between crown temperature and FCVI and NIRv, though the strength of these relationships varied by species. While ōhi‘a had higher NDVI, EVI, and NIRv values than koa, the two species did not differ in FCVI. Species also showed differences in variability among and within crowns. By mapping spatial variability in canopy temperatures this study suggests important species-level differences in sensitivities to climate change, with implications for emergent patterns of tropical forest phenology and productivity at larger extents.