📄 中文摘要
NVIDIA Earth-2开放模型致力于提供从地球系统模拟、数值天气预报到下游应用的全方位气象解决方案。这些模型通过开源方式发布，旨在加速气象科学研究和应用开发。核心内容涵盖了多个关键领域：首先是基础物理模拟，利用高性能计算和AI技术，实现对大气、海洋、陆地和冰盖等地球系统组件的精确建模。其次是数据同化，通过集成卫星、雷达、地面观测等异构数据源，利用AI方法（如深度学习）优化初始条件，显著提高预报准确性。模型还包括了高分辨率次季节和季节尺度预测能力，利用先进的神经网络架构，突破传统物理模型的计算瓶颈，实现更长期、更精细化的天气和气候预测。此外，Earth-2开放模型特别强调对极端天气事件的预测和归因，通过AI增强的模拟和分析工具，帮助科学家更好地理解和预测飓风、洪水、干旱等灾害性天气。在下游应用方面，这些模型支持能源管理、农业生产、城市规划和灾害预警等多个行业，通过提供定制化的气象信息和预测产品，赋能决策者。NVIDIA Earth-2开放模型栈的开放性意味着研究人员和开发者可以自由访问、修改和贡献代码，共同推动气象AI领域的发展，最终目标是构建一个更准确、更实时的全球气象智能平台，以应对气候变化带来的挑战。
📄 English Summary
**NVIDIA Earth-2 Open Models Span the Whole Weather Stack**
NVIDIA Earth-2 Open Models offer a comprehensive meteorological solution spanning from Earth system simulation and numerical weather prediction to downstream applications. These models are released open-source to accelerate meteorological scientific research and application development. The core content covers several key areas: first, fundamental physical simulation, utilizing high-performance computing and AI technologies to achieve precise modeling of Earth system components such as the atmosphere, oceans, land, and ice sheets. Second, data assimilation, integrating heterogeneous data sources like satellites, radar, and ground observations, and employing AI methods (e.g., deep learning) to optimize initial conditions, significantly enhancing forecast accuracy. The models also include high-resolution sub-seasonal and seasonal prediction capabilities, leveraging advanced neural network architectures to overcome computational bottlenecks of traditional physical models, enabling longer-term and more refined weather and climate predictions. Furthermore, Earth-2 open models particularly emphasize the prediction and attribution of extreme weather events, using AI-enhanced simulation and analysis tools to help scientists better understand and forecast disasters such as hurricanes, floods, and droughts. In terms of downstream applications, these models support various industries including energy management, agricultural production, urban planning, and disaster early warning, empowering decision-makers by providing customized meteorological information and forecast products. The open nature of the NVIDIA Earth-2 model stack means researchers and developers can freely access, modify, and contribute code, collectively advancing the field of meteorological AI. The ultimate goal is to build a more accurate and real-time global meteorological intelligence platform to address the challenges posed by climate change.