Abstract
Modern Earth science is at an inflection point. The vast, fragmented, andcomplex nature of Earth system data, coupled with increasingly sophisticatedanalytical demands, creates a significant bottleneck for rapid scientificdiscovery. Here we introduce EarthLink, the first AI agent designed as aninteractive copilot for Earth scientists. It automates the end-to-end researchworkflow, from planning and code generation to multi-scenario analysis. Unlikestatic diagnostic tools, EarthLink can learn from user interaction,continuously refining its capabilities through a dynamic feedback loop. Wevalidated its performance on a number of core scientific tasks of climatechange, ranging from model-observation comparisons to the diagnosis of complexphenomena. In a multi-expert evaluation, EarthLink produced scientificallysound analyses and demonstrated an analytical competency that was rated ascomparable to specific aspects of a human junior researcher's workflow.Additionally, its transparent, auditable workflows and natural languageinterface empower scientists to shift from laborious manual execution tostrategic oversight and hypothesis generation. EarthLink marks a pivotal steptowards an efficient, trustworthy, and collaborative paradigm for Earth systemresearch in an era of accelerating global change. The system is accessible atour website https://earthlink.intern-ai.org.cn.