Predicting solar flares: New study shows how it can be reliably done
A multi-institutional research group based out of China has put together an "early warning" system that could aid in the accurate prediction of space weather, a press release said.
Every 11 years or so, the magnetic field of the Sun flips completely so that its north pole becomes south, while the south pole becomes north. The changes in the magnetic field of the Sun lead to visible changes on the solar surface, where regions of intense magnetic activity temporarily stop the convection process.
The drop in temperature of that region can be observed by telescopes on Earth as they appear darker than the rest of the solar surface and hence are called sunspots. At times, sunspots end up giving out giant eruptions of energy and material, which are called solar flare and coronal mass ejections (CMEs), respectively.
Predicting solar flares
The highly energized eruptions from the Sun pose a major risk to spacecraft as well as infrastructure such as electrical grids. While temporary radio blackouts are common when a solar flare hits Earth's atmosphere, with the sharp rise in the number of satellites being used for communications today, scientists are worried that a major solar storm could lead to a global blackout.
Luckily, the intensity of solar storms we have seen in the past few months has been on the lower end. However, with the Sun now in an active phase of its 11-year solar cycle, the intensity is expected to increase even further. Being able to predict solar flares and the resultant geomagnetic storms that they create can help spacecraft take evasive actions to minimize damage from such events.
How does the prediction technology work?
Scientists have multiple instruments that watch the solar surface for the appearance of sunspots. A collaboration of researchers from multiple institutes in China turned to NASA's Solar Dynamics Observatory (SDO) and dug through its records about sunspots over a period of nine years (2010-2019) to devise a two-stage predictor for solar flares.
The researchers scanned through the SDO data to find entries of solar flare events, their magnitude, and sunspots involved. Then using an unsupervised clustering method, the researchers categorized their datasets and selected the ones with positive samples to be used in the next stage of the algorithm.
Here the researchers turned to neural networks - computer systems that are designed to function like the human brain to predict the likelihood of a solar flare occurring in the next 48 hours. The researchers concluded that using several neural networks improved the prediction performance of their method when compared to using a single neural network.
The research findings have been published in the journal Space: Science and Technology
Solar flares are solar storm events driven by the magnetic field in the solar activity area. Solar flare, often associated with solar proton event or CME, has a negative impact on ratio communication, aviation, and aerospace. Therefore, its forecasting has attracted much attention from the academic community. Due to the limitation of the unbalanced distribution of the observation data, most techniques failed to effectively learn complex magnetic field characteristics, leading to poor forecasting performance. Through the statistical analysis of solar flare magnetic map data observed by SDO/HMI from 2010 to 2019, we find that unsupervised clustering algorithms have high accuracy in identifying the sunspot group in which the positive samples account for the majority. Furthermore, for these identified sunspot groups, the ensemble model that integrates the capability of boosting and convolutional neural network (CNN) achieves high-precision prediction of whether the solar flares will occur in the next 48 hours. Based on the above findings, a two-stage solar flare early warning system is established in this paper. The F1 score of our method is 0.5639, which shows that it is superior to the traditional methods such as logistic regression and support vector machine (SVM).