Bmaina/land-degradation-mapper

Transformer-based geospatial deep learning system for large-scale land degradation mapping using fused Sentinel-1 SAR and Sentinel-2 optical imagery. Supports the African Union Great Green Wall restoration programme. # 🌍 LandDegMapper ### Transformer-Based Land Degradation Mapping via Fused Sentinel-1/2 Imagery --- ## Table of Contents 1. Why This Project Matters 2. The Problem: Land Degradation at Scale 3. Our Solution 4. How the Model Works 5. Why These Specifications 6. Data Sources 7. Model Architecture Deep Dive 8. Training Strategy 9. Degradation Classes 10. Performance 11. Real-World Applications 12. The Great Green Wall Connection 13. Scientific Foundation 14. Installation and Usage 15. Project Structure 16. Colab Notebook 17. Contributing --- ## 1. Why This Project Matters > *"Land degradation affects 3.2 billion people worldwide and costs the global economy > an estimated USD 10.6 trillion annually."* > — IPBES Land Degradation Assessment, 2018 Land degradation — the decline in land productivity, biodiversity, and ecosystem function caused by human activities and climate change — is one of the defining environmental crises of our time. Yet despite its scale, it remains one of the least monitored environmental phenomena. Ground surveys are slow, expensive, and geographically limited. Traditional satellite analysis relies on a single sensor and often misses the full picture. **LandDegMapper changes this.** It provides a scalable, automated, high-resolution pipeline that fuses radar and optical satellite data through a modern AI architecture to map land degradation across entire countries at 10-metre resolution — in minutes rather than months. --- ## 2. The Problem: Land Degradation at Scale ### The Scale of the Crisis ``` ┌─────────────────────────────────────────────────────────────────────┐ │ GLOBAL LAND DEGRADATION FACTS │ ├──────────────────────────────┬──────────────────────────────────────┤ │ Area affected │ 5.2 billion hectares (~38% of land) │ │ People affected │ 3.2 billion │ │ Annual economic cost │ USD 10.6 trillion │ │ Spe …