Airborne tree crown detection for predicting spatial heterogeneity of canopy transpiration in a tropical rainforest

Tropical rainforests comprise high tree species diversity and variable site conditions; their transpiration contributes to climate regulation. At different sites in a lowland rainforest in Sumatra, we tested crown metrics derived from drone-based photogrammetry for predicting and scaling tree transpiration. Studying 42 trees with sap flux measurements, we observed a linear relationship between tree water use and crown surface area (R² = 0.75). Crown metrics explained more of the observed variability in tree water use than the conventionally used stem diameter; thus, uncertainties in stand-scale transpiration estimates substantially decreased. For eight study plots, the number of tree crowns detected by an automatic tree segmentation algorithm (AMS3D) differed only 7%, on average, from ground-based stem counts. Between the four riparian and four upland study plots, scaled stand-level transpiration estimates suggest more than two-fold variability. In conclusion, we regard drone-based crown metrics and canopy segmentation to be very useful tools for the scaling of transpiration from tree to stand level. Our results indicate considerable spatial heterogeneity of transpiration in tropical rainforests.