AI Insight
This paper presents a comprehensive review of Earth science foundation models, which are large AI systems that integrate diverse data types including satellite imagery, climate reanalysis data, geophysical observations, and scientific text to support Earth science applications. The authors organize these models along two dimensions: capability depth (ranging from basic perception to advanced reasoning and scientific discovery) and application breadth (covering atmosphere, oceans, land, biosphere, human systems, ice, and coupled processes). The review catalogs over 200 datasets and benchmarks while identifying critical challenges in data integration, scientific reliability, scalability, and the development of autonomous AI systems for Earth science.
Why it matters
Foundation models represent a paradigm shift in how AI can process and reason about complex Earth system data, potentially accelerating climate research, natural hazard prediction, and environmental monitoring. These systems could enable more integrated approaches to understanding planetary processes and support evidence-based decision-making for environmental policy and resource management.
arXiv:2605.12542v2 Announce Type: replace
Abstract: Large foundation models (FMs) are transforming Earth science by integrating heterogeneous multimodal data, such as multi-platform imagery, gridded reanalysis data, diverse geophysical and geochemical observations, and domain-specific text, to support tasks ranging from basic perception to advanced scientific discovery. This paper provides a unified review of Earth science foundation models (Earth FMs) through two complementary dimensions: depth, which traces the evolution of model capabilities from perception to multimodal reasoning and agentic scientific workflows, and breadth, which summarizes their expanding applications across the atmosphere, hydrosphere, lithosphere, biosphere, anthroposphere, and cryosphere, as well as coupled Earth system processes. Using this framework, we review representative multimodal Earth foundation models and compile more than 200 datasets and benchmarks spanning diverse Earth science tasks and modalities. We further discuss key challenges in multimodal data heterogeneity, scientific reliability and continual updating, scalability and sustainability, and the transition from foundation models to agentic and embodied Earth intelligence, and outline future directions toward more integrated, trustworthy, and actionable AI Earth scientists. Overall, this paper offers a structured roadmap for understanding the development of Earth foundation models from both capability depth and application breadth.
Source: Earth Science Foundation Models: From Perception to Reasoning and Discovery