Estimating individual-level plant traits at scale | JoVE Visualize

Area of Science:

Ecology
Remote Sensing
Computational Biology

Background:

Functional ecology increasingly relies on trait data for understanding ecological communities.
Current methods for trait data collection are limited by either small sample sizes or lack of individual-level resolution.
Bridging this gap requires remote-sensing methods capable of identifying individual trees and estimating their traits.

Purpose of the Study:

To develop and validate a novel pipeline for scaling functional traits from individual trees to landscape levels using remote sensing data.
To integrate structural and leaf-level trait estimations for millions of trees.
To enable large-scale ecological analyses of community composition and ecosystem function.

Main Methods:

Utilized National Ecological Observatory Network (NEON) data, including hyperspectral imagery and field measurements.
Developed a three-stage pipeline: image segmentation for individual tree identification and structural trait estimation, ensemble modeling for inferring leaf traits (LMA, N, C, P) and DBH, and prediction across remote-sensing footprints.
Employed hyperspectral signatures and allometry for trait inference, with an ensemble approach outperforming single models.

Main Results:

Achieved R² values ranging from 0.41 to 0.75 for trait predictions on held-out data, with DBH estimation showing R² of 0.62.
The ensemble modeling approach outperformed single partial least-squares models, though carbon content prediction was less accurate (R²=0.41).
Crown segmentation introduced the most uncertainty, while individual crown-based trait predictions improved accuracy over pixel-based methods by 0.07-0.26 in R².

Conclusions:

The developed pipeline successfully generated individual-level trait data for approximately 5 million trees across 360 km², demonstrating a significant advancement in large-scale ecological data acquisition.
This dataset facilitates landscape-scale testing of ecological hypotheses, revealing correlations between foliar traits, structural traits, and environmental conditions.
The study highlights the potential of integrating remote sensing with advanced analytical methods to revolutionize functional trait research in ecology.