Fig 9: Covariations among 7 traits across all site-species combos. Symbol shape represents species and color represents site. Direction and magnitude of correlations are shown by bubble color (t-test, p < 0.05). Data values are species means
5. CONCLUSION & FUTURE WORK
5.1. CONCLUSION
● NEON Data are useful!
● Trees for the Forest. Crown-level process is essential for forest ecology and our object-based analysis is important for Remote Sensing.
● Tree species significantly differ in crown architectural traits as hypothesized.
● Economic tradeoffs exist at crown scale, not just leaves, expressed by the strong co-variations among the crown traits governing how each species adaptively arranges and orients leaves in their crowns as a strategy to harvest sunlight.
5.2. CONCLUSION
● Greatly expand the current hand-delineated sample by delineating artificial-intelligence (AI) tree crowns using deep learning-based model.
● Expand the data at more NEON sites and other biomes representing wider environmental gradients.
● Refine traits to improve delineation of different aspects of crown architecture. For example, the "shape" of the crown may not simply be described by top- or bottom-heaviness due to the multi-modal distribution of many trees. Thus, APAD50 alone is not sufficient enough to represent a crown shape.
6. CROWN TRAITS ILLUSTRATION