Researchers have introduced TerraNova, a novel foundation model designed to integrate physical Earth data with societal indicators. Unlike previous models that averaged data across political borders, TerraNova processes gridded Earth-system fields and national indicators in their native geometries. It utilizes dedicated encoders for location, country, time, and task, fused by cross-modal transformers into a shared spatiotemporal state. The model employs contrastive objectives to align country-level data with geographic coordinates and pretrained geospatial embeddings, enabling it to perform competitively on geospatial tasks while also supporting country-level capabilities. AI
IMPACT This model could enable more comprehensive analysis of coupled Earth and human systems, potentially improving predictions and policy-making.
RANK_REASON The cluster contains a research paper detailing a new foundation model. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
