From pixels to planning: Earth AI for nature restoration

2026-06-16 · Source: The latest research from Google · Field: Science & Research — Environmental Science & Earth Systems, Mathematics & Computational Sciences · Depth: Intermediate, short

Summary

Google Research, in collaboration with the Leverhulme Centre for Nature Recovery at the University of Oxford, has released a new vectorized dataset transforming high-resolution maps into an actionable inventory of fine-scale ecological features like hedgerows, stone walls, and copses across the UK. This builds upon their previous Farmscapes 2020 raster map, which identified overlooked features in England. The new deep learning framework overcomes challenges in spatial topology, semantic classification, and computational scale. It utilizes a Remote Sensing Foundations' (RSF) Vision-Transformer (ViT) Backbone, pre-trained on over 300 million global satellite images, fine-tuned with ~247 km² of annotated data. A dual-layer labeling system with submeter imagery and 1-meter LiDAR data, a scalable algorithm for merging geometries across S2-cell tiles, and the Polsby–Popper compactness score (e.g., <0.5 for linear features) are employed. Processing millions of features across 130,000 km² of England is achieved using Google Earth Engine. This resource aims to empower landowners and conservationists to measure and expand these features for nature restoration without compromising food security.

Key takeaway

For conservationists and landowners planning nature restoration projects, you should utilize the new vectorized dataset to precisely identify and quantify fine-scale ecological features like hedgerows and copses. This resource enables accurate carbon accounting and biodiversity enhancement on working lands, ensuring your efforts do not compromise food security. Consider integrating this data to detect potential "leakage" and optimize restoration strategies across your land.

Key insights

High-resolution deep learning can reveal fine-scale ecological features for nature restoration and carbon accounting.

Principles

• Fine-scale woody features enhance carbon and biodiversity without displacing crops.
• Overlapping spatial features require dual-layer mapping.
• Semantic classification needs geometric intelligence.

Method

A deep learning framework uses a pre-trained ViT backbone, fine-tuned with LiDAR and submeter imagery, then applies a dual-layer labeling system, geometry merging, and Polsby–Popper compactness scores on Google Earth Engine.

In practice

• Identify hedgerows, stone walls, and copses for restoration.
• Quantify fine-scale woody features in silvopasture systems.
• Detect "leakage" events in conservation projects.

Topics

• Deep Learning
• Geospatial Data
• Nature Restoration
• Farmscapes
• Google Earth Engine
• Biodiversity Conservation

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• AI4Land: Scalable Deep Learning for Global High-Resolution Land Use Reconstruction — AI4Land uses deep learning and HPC to reconstruct global high-resolution land use for improved climate modeling.
• Degradation-based augmented training for robust individual animal re-identification — Augmenting training data with diverse degradations significantly improves wildlife re-identification robustness against real-world image quality issues.
• 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.
• An Improved Generative Adversarial Network for Micro-Resistivity Imaging Logging Restoration — A specialized GAN architecture significantly improves micro-resistivity imaging log restoration by integrating multi-scale feature learning and dual discrimination.
• Improved Knowledge Distillation for Land-Use Image Classification — Combining hard and soft supervision with KL divergence and Cosine Similarity improves knowledge distillation for land-use classification.
• How we’re helping preserve the genetic information of endangered species with AI — AI tools dramatically accelerate and reduce the cost of genome sequencing for endangered species conservation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The latest research from Google.