Self-Supervised Tree-level Biomass Estimation in Urban Environments From Airborne LiDAR and Optical Observations

2026-06-24 · Source: Computer Vision and Pattern Recognition · Field: Science & Research — Environmental Science & Earth Systems, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

A novel framework quantifies crown-level above-ground biomass (AGB) for an 810~km$^2$ urban landscape in Ontario, Canada, utilizing 2018 and 2023 leaf-off airborne LiDAR (8--10~pulses~m$^{-2}$) and near-infrared RGB orthophotography (0.16--0.20~m). This system employs a dual-stream cross-attention network, trained with rule-based pseudo-labels, to semantically mark buildings, needleleaf, and deciduous trees, achieving 0.86 precision, 0.83 recall, and 0.84 Dice scores on independent tiles. Multiscale watershed segmentation delineates crowns, with AGB estimated via a crown area--height power-law proxy calibrated to species-specific allometry (Lambert et al., 2005) for 21,921 inventory trees. AGB prediction yielded R^2=0.609 with inventory geometry and R^2=0.570 operationally, highlighting crown delineation as the primary uncertainty. Aggregated 30m estimates show 1.73~Tg AGB in 2018 and 1.81~Tg in 2023, representing a 39~Gg~C net gain over five years. The framework operates without manual annotation, using standard provincial data, and generates a public bitemporal crown-level AGB database.

Key takeaway

For urban planners and environmental scientists seeking scalable, accurate urban tree biomass quantification, this framework offers a robust, annotation-free solution. You can utilize standard provincial LiDAR and optical data to generate bitemporal crown-level AGB maps, enabling precise carbon stock assessments and monitoring. Focus on refining crown delineation techniques to further reduce estimation uncertainty.

Key insights

Self-supervised deep learning combined with LiDAR and orthophotography enables accurate, scalable urban tree biomass estimation.

Principles

• Pseudo-labeling can train semantic segmentation networks effectively.
• Crown delineation is a key uncertainty in AGB estimation.
• Bitemporal data reveals urban carbon stock changes.

Method

A dual-stream cross-attention network, trained on rule-based pseudo-labels, performs semantic segmentation. Multiscale watershed segmentation delineates crowns, followed by AGB estimation using a calibrated crown area-height power-law proxy.

In practice

• Use LiDAR and orthophotography for urban tree mapping.
• Implement pseudo-labeling to avoid manual annotation.
• Monitor urban carbon sequestration with bitemporal AGB maps.

Topics

• Self-Supervised Learning
• Airborne LiDAR
• Urban Biomass Estimation
• Semantic Segmentation
• Crown Delineation
• Carbon Sequestration

Best for: AI Scientist, Research Scientist

Related on AIssential

• SegmentAnyTreeV2: Scaling Transformer-Based Tree Instance Segmentation Across Sensors, Platforms, and Forests — SegmentAnyTreeV2 offers robust, scalable tree instance segmentation for forest point clouds using a Transformer-based architecture.
• Aerial-ground LiDAR place recognition with patch-level self-supervised learning and expanded reciprocal re-ranking — Aerial-ground LiDAR place recognition is improved by a novel framework combining patch-level self-supervised learning and an Expanded Reciprocal re-ranking algorithm.
• Integrating national forest inventory, airborne lidar, and satellite imagery for wall-to-wall mapping of forest structure with computer vision — Integrating lidar and satellite data with computer vision improves large-scale forest attribute mapping accuracy and consistency.
• Colour Extraction Pipeline for Odonates using Computer Vision — A computer vision pipeline segments Odonates and extracts color data to enable large-scale ecological correlation analysis.
• GMBFormer: An NDVI-Guided Global Memory Bank Transformer for Urban Green-Space Extraction from Ultra-High-Resolution Imagery — GMBFormer improves urban green-space extraction by decoupling NDVI and using a global memory bank for semantic reuse.
• Contrasting Local and Global Modeling with Machine Learning and Satellite Data: A Case Study Estimating Tree Canopy Height in African Savannas — Local SatML models can outperform global or fine-tuned global models for specific regional tasks.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.