(DURIP) SPACE WEATHER FORECASTING TESTBED FOR MACHINE LEARNING RESEARCH AND EDUCATION
Abstract
We propose to acquire machine learning Graphics Processing Unit (GPU) servers, an associated large-scale on-site data storage array, and a video display wall to equip the new University of Colorado at Boulder Space Weather Deep Learning Laboratory (DLL). These elements will combine into a system enabling major advances in machine learning research applied to space weather forecasting. The proposed system will also enable the first space weather Testbed capability in an academic research facility, accelerating the Research-to-Operations (R2O) transition of new prediction and visualization tools to operational space weather forecasting centers such as the USAF 557th Weather Wing in Omaha, Nebraska. The resulting DLL environment will enable undergraduates, graduate students, and post-docs access to state-of-the-art machine learning tools and a simulated forecasting environment in the Testbed that will provide a unique education in the field of operational space weather forecasting. The main elements of the proposed equipment include one NVIDIA DGX-A100 rack-mounted machine learning GPU server (8 Telsa A100 GPUs, 320 GB GPU memory, 5 PetaFLOP performance), a DDN AI200X 88TB storage array with a Mellanox InfiBand 100Gb-sec connection to the DGX-A100, and a Barco UniSee 800 2x2 LCD video wall (4x1920x1080 800 nits HDTV bezel-less screens) and ancillary computer systems for screen management and visualization tool development. The capabilities of the proposed DURIP equipment will significantly advance our first-of-its-kind machine learning laboratory and space weather forecasting testbed, accelerating our current programs in developing GNSS scintillation prediction, solar flare onset prediction, radiation belt enhancement, LEO satellite drag prediction, and Geomagnetically Induced Current (GIC) events and enabling hands-on student involvement in operational forecasting research. The enabled research will directly contribute to the national effort to improve forecasting, nowcasting, and mitigation of severe to extreme space weather impacts on DoD and civil critical infrastructure and operational missions.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 07, 2023
- Source ID
- FA95502110267
Entities
People
- Thomas Berger
Organizations
- Air Force Office of Scientific Research
- Regents of the University of Colorado
- United States Air Force