Imaging the Solar Interior: Emerging
Active Regions

Overview

In September 1859, the Sun released a tremendous amount of energy in the form of charged particles and electromagnetic waves in what is now called a Coronal Mass Ejection (CME). Roughly 17 hours after its initial release, the CME impacted Earth’s magnetosphere with enough intensity that it caused most telegraph systems to fail—in some areas it was reported that the telegraph lines themselves were sparking—and the aurora was visible in the Northern Hemisphere as far south as the Caribbean. This was the so-called Carrington event—the largest geomagnetic storm on record. If such an event were to occur today, it would almost certainly overload the power grid on the day side of the Earth, causing billions of dollars in damage. Luckily, our infrastructure is designed to avoid most of the damage of such a strong CME—if there is enough time to prepare these systems.

Our work is focused on building the groundwork for such a warning system. We use velocity data from NASA’s Solar Dynamics Observatory (SDO) to compute images of the solar interior, similar to the way ultrasounds map our bodies based on the travel time of sound waves. Using these images, we observe time-dependent structures that we interpret as rising magnetic flux, which eventually forms active regions on the Sun’s surface. These active regions are hotbeds for solar flares and CMEs, so tracking their formation is the first step in building a reliable space weather forecasting tool.

Project Details

The Helioseismic and Magnetic Imager (HMI) aboard SDO captures images of the Sun and produces maps of the line-of-sight velocity, among other data products. We have previously developed an algorithm which uses these velocity maps, or dopplergrams, to produce the images of the solar interior. Specifically, we focus on regions between 40,000 and 70,000 kilometers from the surface of the Sun, because the heliophysics community has hypothesized that magnetic structures form in this area. The images we obtain show how quickly or slowly sound waves travel through a particular area, and we are currently developing a technique to associate these changes to local conditions, such as density and magnitude of the magnetic field.

Results and Impact

Recently, we completed a proof-of-concept study to demonstrate the sensitivity of our imaging method, and we are currently working on a larger-scale statistical study of historical active regions. This includes studying past data records of active regions ranging from the mid-1990s to today so we can gain a better understanding of these structures. We aim to produce a machine learning algorithm, trained on the images of these past active regions, that can quickly and accurately identify which regions of the Sun are most likely to be active. A large majority of solar flares originate in these active regions, so understanding their behavior and properties is key to developing a space weather forecasting tool.

Why HPC Matters

NASA’s high-performance computing resources enable us to process a six-hour, 100-by-100 pixel dataset in about four minutes, using 1,000 processors on the Pleiades supercomputer, along with code optimization techniques. While four minutes doesn’t sound like very much, we process up to 30 images in the study of one active region. If we were to process the images for one active region on a standard home computer, it would take three months!