FLORA: A deep learning approach to predict forest attributes from heterogeneous LiDAR data

2026-06-30 · Source: Artificial Intelligence · Field: Science & Research — Environmental Science & Earth Systems, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

FLORA (Forest LiDAR Octree Regression with Auxiliary Data) is a deep learning framework designed to predict six critical forest attributes: dominant height, total volume, deciduous volume, coniferous volume, basal area, and stem density. It addresses the challenge of heterogeneous LiDAR data, which arises from varied sensors, flight parameters, seasons, and scan angles in national LiDAR programs. FLORA integrates an octree-based backbone with ecological and spatiotemporal auxiliary variables through a late-fusion gating mechanism. Trained and evaluated on 32,052 National Forest Inventory plots across mainland France using data from the French LiDAR HD program, a single FLORA model trained on both leaf-on and leaf-off acquisitions outperformed season-specific models, enhancing cross-season robustness. Auxiliary variables provided modest overall gains but significantly improved species-specific volume prediction. FLORA achieved an rRMSE of approximately 12.3% (R2 = 0.88) for dominant height and 39% (R2 = 0.74) for total volume, establishing a robust baseline for large-scale forest attribute estimation.

Key takeaway

For Machine Learning Engineers developing large-scale forest attribute prediction models, you should adopt FLORA's deep learning architecture. Its octree-based backbone, combined with a late-fusion gating mechanism for ecological and spatiotemporal auxiliary variables, offers robust performance across heterogeneous LiDAR data. Integrating both leaf-on and leaf-off acquisitions will significantly improve cross-season model robustness. This approach provides a strong baseline for national forest inventory estimates, achieving high accuracy for dominant height and total volume.

Key insights

FLORA is a deep learning framework predicting forest attributes from heterogeneous LiDAR data using an octree backbone and auxiliary variables.

Principles

• Heterogeneous LiDAR data requires robust models beyond local calibration.
• Combining leaf-on and leaf-off data improves cross-season model robustness.
• Auxiliary variables enhance species-specific volume prediction.

Method

FLORA combines an octree-based deep learning backbone with ecological and spatiotemporal auxiliary variables via a late-fusion gating mechanism to predict six forest attributes.

In practice

• Integrate diverse LiDAR acquisitions (leaf-on/off) for broader model applicability.
• Incorporate ecological and spatiotemporal data for improved species-specific predictions.

Topics

• FLORA Framework
• Deep Learning
• LiDAR Data Analysis
• Forest Inventory
• Octree Regression
• Auxiliary Variables

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• STELLAR: Spatio-Temporal Environmental Learning with Latent Alignment and Refinement for Long-Tailed Species Distribution Modeling — STELLAR jointly optimizes dynamic habitat context and community structure in a shared latent space for improved species distribution modeling.
• Comparison of Deep Learning Frameworks For Rice Disease Mapping From UAV Multispectral Imaging — Lightweight CNNs excel in UAV-based rice disease mapping, with U-Net++ EfficientNet-B3 achieving 97.62% mIoU.
• Is XGBoost gone: How Relational Foundation Models Conquered 500 Billion Row Enterprise Data — The dominance of gradient boosted trees on tabular data was an illusion created by pre-processed Kaggle datasets.
• From pixels to planning: Earth AI for nature restoration — High-resolution deep learning can reveal fine-scale ecological features for nature restoration and carbon accounting.
• Random Forests as Statistical Procedures: Design, Variance, and Dependence — Random forests are finite-sample statistical designs with irreducible predictive variability due to inherent design-induced dependence.
• Learning Order Forest for Qualitative-Attribute Data Clustering — This paper introduces "Clustering with Order Forest learning" (COForest), a novel paradigm addressing the challenge of clustering qualitative (categorical/nominal) attribute data where traditional Euc…

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.