Search

Earth-2 is a collaborative and open-science initiative launched by NVIDIA in Nov 2021, to develop operational digital twins of the Earth and its subsystems. The primary aim of Earth-2 is to improve predictions of extreme weather, projections of climate change, and to accelerate the development of effective mitigation and adaptation strategies - all using advanced modeling and scientifically principled machine learning methods at large scale. An AI centric approach is critical to enable fast and interactive exploration of potential adaptation / mitigation strategies, their consequences, and their trade-offs. Some of the digital twins under development as part of this project include AI surrogates of medium-range weather, global-climate, and seasonal-to-subseasonal forecasting. Other efforts include generative AI downscaling, automatic discovery of physical parameterizations, and generative prediction of precipitation. This talk will discuss the goals, strategies, progress, applications, and future challenges of this effort., The 28th IUGG General Assembly (IUGG2023) (Berlin 2023)