Totality! Predicting the Corona of the 2017 Great American Solar Eclipse
Overview
The solar corona—the region extending from the Sun’s surface to several million kilometers out into space—has a highly complicated structure and displays a multitude of physical processes at various spatial and temporal scales. Due to this inherent complexity, understanding the physics of the corona represents a challenging problem theoretically and computationally.
By bringing together modern advances in computational infrastructure, physical models, and real-time observational capabilities, we can run large simulations to reproduce and predict many of the features of the corona. Total solar eclipses offer a unique opportunity to validate current models, as they provide a pristine, unobstructed view of the entire corona from Earth.
Besides pure scientific interest, understanding and predicting the coronal structure has important applications in forecasting space weather, which is relevant for space travel, as well as air, ground, and satellite operations.
Project Details
Our model uses numerical realizations of resistive, thermodynamic, magnetohydrodynamic (MHD) equations to simulate the global corona, with observations of the solar surface magnetic field as a boundary condition. For this simulation, we implemented new methodologies including a wave-turbulence-driven coronal heating model, and the energization of the magnetic field via inserted flux ropes along filament channels.
In order to capture the fine details of the corona, we ran the model using one of the largest computational grids for simulations of this type. We numerically integrated the MHD equations until the system came to a near steady state. We then generated synthetic observations and other images that could be compared to eclipse photographs.
Results and Impact
The 2017 Great American Solar Eclipse offered us an opportunity to test state-of-the-art coronal modeling, which resulted in a good agreement with observations. In fact, our predictions were used for observational planning purposes prior to the eclipse by some NASA-involved experiments.
Due to the lack of far-side observational capabilities, the solar surface magnetic field data used in our simulation was limited to data obtained from an Earth-Sun view. This resulted in some areas of data being several weeks old, as the Sun’s rotation period is 28 days; structures in our simulations that do not match as well with observations tend to originate in these areas. Expanded observational coverage of the far side of the Sun by future satellite missions would greatly improve our predictive capability.
Currently, we are working with other scientists to quantitatively compare our simulation with eclipse observations. This collaboration not only helps us improve our models, but analysis of the three-dimensional simulation data is crucial for interpreting complex coronal observations.
Why HPC Matters
The final high-resolution simulation of the solar corona used a grid of 295 x 315 x 699 (65 million cells), and simulated a total of approximately 60 physical hours. A simulation of this size could not have been achieved without substantial HPC resources. We performed the run using the Pleiades system at the NASA Advanced Supercomputing (NAS) Division, using 210 Intel Ivy Bridge nodes (4,200 processors). It took about 60 wall-clock hours to complete—about as long as the time period simulated. This ability to run in near-real-time paves the way for continually running models that routinely ingest observations—a feat that would highly benefit the space weather and solar modeling communities.
More Information
- Spoiler Alert: Computer Simulations Provide Preview of Upcoming Eclipse
- See additional images/videos related to this demo in our media gallery.
- See the poster for this demonstration (PDF, 3.4MB), as shown in the NASA booth at SC17.
caplanr@predsci.com Cooper Downs, Predictive Science Inc.
cdowns@predsci.com
