Geospatial Foundation Models Benchmark for Biomass Estimation

A new research paper explores the effectiveness of Geospatial Foundation Models (GFMs) for estimating above-ground biomass (AGB) from satellite imagery. The study benchmarks 11 GFMs using the AGBD dataset, comparing models run as frozen encoders against pre-computed embedding products like AlphaEarth Foundations (AEF) and TESSERA. Results indicate that while frozen encoders underperform, pre-computed embeddings, particularly AEF, significantly enhance biomass estimation accuracy, with an MLP trained on AEF embeddings outperforming the supervised state-of-the-art. AI

IMPACT This research could improve the accuracy and scalability of biomass estimation, crucial for climate change monitoring and carbon stock management.

RANK_REASON Research paper detailing a benchmark of foundation models for a specific scientific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• above-ground biomass
• AEF
• AGBD dataset
• AlphaEarth Foundations
• ESA CCI Soil Moisture for improved Earth system understanding: State-of-the art and future directions
• Geospatial Foundation Models
• Ghjulia Sialelli
• Hugging Face
• Pangaea
• TESSERA

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →