Curated News
By: NewsRamp Editorial Staff
November 03, 2026
Satellite AI Counts 5.25 Billion Trees in East Africa
TLDR
- University of Copenhagen's deep-learning framework outpaces ResNet50, SwinT, and ViT, enabling scalable tree counts for restoration advantage.
- The framework uses a hybrid UNet-ResNet50 on 20-band multisensor satellite patches to regress tree counts, calibrated with national forest inventories.
- This scalable baseline helps governments track restoration, guide planting, and account for trees beyond forests, improving global reforestation outcomes.
- The study counted 5.25 billion trees in Rwanda and Tanzania, revealing that over half grow outside areas classified as tree cover.
Impact - Why it Matters
This news matters because global tree-planting pledges have surged since the Paris Agreement, yet reliable, scalable baselines for counting trees have been lacking. Field inventories are costly and spatially limited, and coarse vegetation-index approaches miss trees outside forests. By providing a repeatable, satellite-based method that captures more than half of trees in nonforest landscapes, this framework gives governments and restoration programs a practical tool to plan planting, monitor gains and losses, and allocate resources more effectively. It also challenges the conventional forest-centric view of tree cover, which could reshape how restoration targets are set and verified across diverse ecosystems.
Summary
A research team led by the University of Copenhagen has developed a deep-learning framework that combines multisensor satellite imagery to estimate tree abundance across heterogeneous landscapes. published (DOI: 10.34133/remotesensing.0970) the study on August 18, 2026, in the Journal of Remote Sensing. The framework integrates PlanetScope, Sentinel-1, and Sentinel-2 data on a common 3-meter grid and uses a hybrid UNet-ResNet50 architecture to predict tree counts from 96 × 96 meter image patches. Instead of detecting individual crowns, it infers tree numbers from spatial, spectral, and textural patterns, avoiding reliance on explicit crown boundaries that confound object-detection methods in dense or nonforest systems.
Applied to Rwanda and Tanzania, the model estimated approximately 5.25 billion trees—about 213 million in Rwanda and 5.04 billion in Tanzania after calibration against National Forest Inventory data. Only 48.8% of these trees fell within the WorldCover "tree cover" class; the remaining 51.2% occurred across shrublands, grasslands, croplands, built-up areas, wetlands, sparse vegetation, and mangroves. The UNet-R50 architecture achieved a mean absolute error of 17.32 and an r² of 0.73, outperforming ResNet50, ResNeXt50, SwinT, and ViT. Sensor-ablation tests showed that Sentinel-1 plus Sentinel-2 alone reached an MAE of 18.42 and r² of 0.77, comparable to the three-sensor composite. Training used about 12 million tree-crown centers from 2,832 sample areas spanning roughly 608,483 hectares, with labels derived from manual interpretation of submeter imagery and semi-automated extraction from airborne and UAV LiDAR canopy-height models.
The researchers emphasize the framework's practical value, noting it "can guide the planning and optimization of resource allocation for tree restoration efforts at national scales." Repeatable satellite-based baselines could help governments track restoration outcomes, identify appropriate planting areas, and account for trees beyond conventional forest boundaries. Future work aims to improve performance in dense closed-canopy forests and incorporate understory or smaller woody plants, while exploring attention-based sensor fusion, domain adaptation, and transfer learning to extend the approach to new regions and sensors.
Source Statement
This curated news summary relied on content distributed by 24-7 Press Release. Read the original source here, Satellite AI Counts 5.25 Billion Trees in East Africa
