The Evolution of GEOS as a Digital Twin for Earth System Science
Overview
The NASA Global Earth Observing System (GEOS) model and assimilation system is NASA’s flagship system for enhancing the use of the agency’s extensive Earth observations. GEOS enables forefront modeling studies that support the planning of future observation missions and advanced data assimilation studies, maximizing the impacts of NASA data for use in weather prediction, climate research, oceanic and atmospheric circulation, and other fields of study. NASA Goddard Space Flight Center’s Global Modeling and Assimilation Office (GMAO) is using historical observations to produce digital models of the Earth system’s climate, which has long been a goal at the center of GEOS development. With the vast expansion of machine learning capabilities and improved programming paradigms for graphics processing units (GPUs), GEOS is now poised to provide an experimental framework within NASA for the creation of true Earth system digital twins for weather and climate.
Project Details
The GEOS framework is ideal for Earth digital twin creation due to its flexibility: it supports both high temporal and 3D physical resolution as well as the exploration of increasingly complex Earth systems such as advanced carbon cycles, global aerosols, ocean dynamics, and biogeochemistry. As such, a digital twin for carbon studies represents a unique configuration of GEOS as compared to a digital twin for severe storms and tornados. The balance between complexity, resolution, and computing capability drives the application of GEOS to the production of Earth digital twins, which can play roles in improving the use of NASA observations or defining new Earth observing system missions.
Artificial intelligence (AI) is being integrated into GEOS chemistry components to enable improved and faster simulations of these complex Earth system processes. GMAO is exploring use of AI for replacing complex physics parameterizations for clouds, aerosols, and radiation processes. The entire GEOS code base is being refactored to introduce a domain-specific language to optimize the use of GPUs on computationally intense portions of GEOS, including the dynamics and transport of aerosols. These advances within GEOS demonstrate the continued evolution of GMAO’s Earth digital twin capability as NASA explores new computing platforms and tackles the challenges of producing and analyzing petabytes of data for research and development.
Results and Impact
GMAO will highlight the advances in data assimilation capabilities and the creation of Earth digital twins with GEOS across a range of capabilities including: coupled ocean-atmosphere Earth system modeling, advanced studies of carbon emission and transport at ultra-high resolutions, and pushing the edge of compute capability in the creation of a 1.5-kilometer-resolution, global digital twin.
Why HPC Matters
The NASA Center for Climate Simulation (NCCS) and the NASA Advanced Supercomputing (NAS) Division provide essential high-end computing (HEC) resources for preparing GEOS modeling systems for emerging HEC systems and exploring new science developments.
What's Next
While current GEOS configurations require thousands of computational cores, configurations over the coming decade will require scalability approaching millions of cores. NASA’s GMAO will continue to prepare its Earth system models for future HEC platforms by exposing new depths of parallelism within existing GEOS application components.
More Information
Researchers
- William M. Putman, NASA Goddard Space Flight Center Please enable Javascript to see email address.