Google AlphaEarth Embeddings Show Promise for Landslide Mapping

A new study published on arXiv evaluates Google AlphaEarth embeddings for landslide susceptibility mapping (LSM), comparing them against traditional landslide conditioning factors (LCFs). The research utilized three deep learning models—CNN1D, CNN2D, and Vision Transformer—across three distinct geographical regions. Results indicate that AlphaEarth embeddings consistently outperformed LCFs, leading to higher accuracy and more stable error distributions in susceptibility maps. AI

IMPACT Google AlphaEarth embeddings show potential as a standardized, information-rich alternative for geospatial analysis tasks like landslide susceptibility mapping.

RANK_REASON Research paper published on arXiv detailing the evaluation of a new geospatial embedding model for a specific application. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• AE embeddings
• CNN1D
• CNN2D
• Emilia-Romagna
• Google AlphaEarth
• Hong Kong
• Italy
• landslide conditioning factors
• Nantou County
• Qinfeng Zhu
• Taiwan
• vision transformer

• paper
• product

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