Earth Science Foundation Models Integrate Multimodal Data for Discovery

This paper explores the development and application of Earth science foundation models (Earth FMs), which leverage large AI models to integrate diverse datasets for Earth observation and analysis. The review covers the evolution of model capabilities, from basic perception to complex reasoning and scientific workflows, and their applications across various Earth systems, including the atmosphere, hydrosphere, and biosphere. Key challenges include managing data heterogeneity, ensuring scientific reliability, and achieving scalability. The transition from foundation models to agentic Earth intelligence is also discussed, highlighting the potential for AI to autonomously conduct scientific investigations. The compilation of over 200 datasets and benchmarks provides a valuable resource for researchers in the field. The development of robust and reliable Earth FMs is crucial for addressing pressing environmental challenges and improving our understanding of Earth system dynamics. The paper offers a roadmap for future research, emphasizing the need for integrated, trustworthy, and actionable AI solutions for Earth science.

_Context: This intelligence report was compiled by the DailyOrbitalWire Strategy Engine. Verified for Art. 50 Compliance._