Improved and Interpretable Solar Flare Predictions With Spatial and Topological Features of the Polarity Inversion Line Masked Magnetograms

Abstract

Many current research efforts undertake the solar flare classification task using the Space‐weather HMI Active Region Patch (SHARP) parameters as the predictors. The SHARP parameters are scalar quantities based on spatial average or integration of physical quantities derived from the vector magnetic field, which loses information of the two‐dimensional spatial distribution of the field and related quantities. In this paper, we construct two new sets of spatial features to expand the feature set used for the flare classification task. The first set uses the idea of topological data analysis to summarize the geometric information of the distributions of various SHARP quantities across active regions. The second set utilizes tools coming from spatial statistics to analyze the vertical magnetic field component Br and summarize its spatial variations and clustering patterns. All features are constructed within regions near the polarity inversion lines (PILs) and classification performances using the new features are compared against those using SHARP parameters (also along the PIL). We found that using the new features can improve the skill scores of the flare classification model and new features tend to have higher feature importance, especially the spatial statistics features. This potentially suggests that even using a single magnetic field component, Br, instead of all SHARP parameters, one can still derive strongly predictive features for flare classification.Plain Language SummaryOur research is targeted at improving the accuracy of solar flare classification by training machine learning models with new interpretable features beyond well‐known physics‐based predictors. We count the number of closed loops and calculate multiple summary statistics of the spatial distribution of high‐resolution magnetic field images of solar active regions to boost the classification result of strong and weak flares. Our results reveal that the spatial distribution of local physical quantities derived from the magnetograms, beyond those commonly adopted, aggregated quantities, can be helpful to improve flare predictability.Key PointsWe adopt a polarity inversion line (PIL) detecting algorithm to obtain PIL masks for the Br component and several SHARP parameter mapsWe construct two sets of spatial statistics features and a collection of topological features based on the PIL masksOur newly constructed features, by itself or joint with topological features, can significantly improve flare predictions than using SHARP parameters only