NG004-0034
SOLSTICE: Space Weather Modeling Meets Machine Learning
Tuesday, 15 December 2020
Poster
Tamas I Gombosi1, Natalia Y Ganushkina1, Spiro K Antiochos2, Yang Chen3, Alfred O Hero4, David Fouhey5, Enrico Landi6, K D Leka7, Yang Liu8, Ward Manchester9, Gabor Toth10, Shasha Zou9 and SOLSTICE Team, (1)University of Michigan Ann Arbor, Ann Arbor, MI, United States, (2)NASA GSFC, Silver Spring, MD, United States, (3)University of Michigan, Department of Statistics, Ann Arbor, MI, United States, (4)University of Michigan Ann Arbor, Ann Arbor, United States, (5)University of Michigan Ann Arbor, Computer Science, Ann Arbor, MI, United States, (6)University of Michigan, Ann Arbor, MI, United States, (7)NorthWest Research Associates, Boulder, CO, United States, (8)Stanford University, HEPL, Stanford, CA, United States, (9)University of Michigan, Climate and Space Sciences and Engineering, Ann Arbor, MI, United States, (10)University of Michigan, Department of Climate and Space, Center for Space Environment Modeling, Ann Arbor, MI, United States
Abstract:
The last decade has truly witnessed the rise of the machine age. The enormous expansion of technology that can generate and manipulate massive amounts of information has transformed all aspects of society. Missions such as SDO and MMS, and numerical models such as the Space Weather Modeling Framework (SWMF) are now routinely generating terabytes of science data, far beyond what can be analyzed directly by humans. Fortunately, concurrent with this explosion in information has come the development of powerful capabilities, such as machine learning (ML) and artificial intelligence (AI), that can retrieve revolutionary new understanding and utility from the massive data sets.
SOLSTICE (Solar Storms and Terrestrial Impacts Center) is a recently selected NASA DRIVE Science Center. It will serve as the vanguard for developing and applying ML methods, which will then raise the capabilities of the entire community. We will combine next generation ML technology with our world-class numerical models and the exquisite data from the space missions to make breakthrough advances in Heliophysics understanding and space weather capabilities. Following that, we will transition our technology to the CCMC for the benefit of all and make our results available to operational organizations for consideration for use in space weather operations.
We describe recent advances made by SOLSTICE on three Grand Challenge Problems:
(i) identifying the onset mechanism of solar flares and coronal mass ejections using interpretable deep learning models, archived solar observations and high-performance physics-based simulations;
(ii) predicting solar storms many hours before eruption use high cadence observations and physics-based feature learning; and
(iii) training time-to-event models to predict event times and flare magnitudes using innovative machine learning techniques.
