Characterizing the Shape of Solar Flares using Topological Data Analysis

Solar flares are eruptions of high energy radiation from the sun, which if powerful enough, can cause widespread blackouts and loss of communication. Prediction of these flares is essential to protect Earth’s electricity-based infrastructure.

The Atmospheric Imaging Assembly collects images of the Sun, taking images in seven different wavelengths – 94 Å, 131 Å, 171 Å, 193 Å, 211 Å, 304 Å, and 335 Å – repeating the cycle with a cadence of every 12 seconds. Flares are categorized by increasing intensity starting from A and going through B, C, M, and X classes. The Geostationary Operational Environmental Satellite has recorded solar emissions and their intensities for over a decade and using this data we are able to place flares into their classes.

Recently, there has been more interest in using Machine Learning (ML) techniques to predict occurrences of solar flares. ML based models are used to correlate various features extracted from solar flare data. We propose to apply Topological Data Analysis (TDA), a mathematical approach that studies the structures and shapes of datasets.

In this talk, we present a TDA-based approach to extract topological features of AIA images and topological features of flare intensity time series and discuss possible application for flare prediction problems.