Forecasting SEP Events During Solar Cycles 23 and 24 Using Interpretable Machine Learning

TL;DR

This study utilizes the recently developed dataset that combines the Solar Dynamics Observatory/Helioseismic and Magnetic Imager's (SDO/HMI) Space weather HMI Active Region Patches (SHARP) and the Solar and Heliospheric Observatory/Michelson Doppler Imager's (SoHO/MDI) Space Weather MDI Active Region Patches (SMARP) to evaluate the predictive potential.

Abstract

Prediction of the Solar Energetic Particle (SEP) events garner increasing interest as space missions extend beyond Earth's protective magnetosphere. These events, which are, in most cases, products of magnetic reconnection-driven processes during solar flares or fast coronal-mass-ejection-driven shock waves, pose significant radiation hazards to aviation, space-based electronics, and particularly, space exploration. In this work, we utilize the recently developed dataset that combines the Solar Dynamics Observatory/Helioseismic and Magnetic Imager's (SDO/HMI) Space weather HMI Active Region Patches (SHARP) and the Solar and Heliospheric Observatory/Michelson Doppler Imager's (SoHO/MDI) Space Weather MDI Active Region Patches (SMARP). We employ a suite of machine learning strategies, including Support Vector Machines (SVM) and regression models, to evaluate the predictive potential of this new data product for a forecast of post-solar flare SEP events. Our study indicates that despite the augmented volume of data, the prediction accuracy reaches 0.7 +- 0.1, which aligns with but does not exceed these published benchmarks. A linear SVM model with training and testing configurations that mimic an operational setting (positive-negative imbalance) reveals a slight increase (+ 0.04 +- 0.05) in the accuracy of a 14-hour SEP forecast compared to previous studies. This outcome emphasizes the imperative for more sophisticated, physics-informed models to better understand the underlying processes leading to SEP events.

Figures (7)

• Figure 1: Probability density distribution of the 3356 positive (green) and negative (red) events in our dataset that occurred between April 4, 1996, and May 9, 2023 (SHARP-SMARP dataset availability) for the logarithm of flare intensities (left panel) and angular distances to the magnetic footpoint of the Earth (right panel).
• Figure 2: Scheme of a workflow to prepare ML-ready datasets from the SMARP and SHARP bobra2021smarpskosovich2024time. The arrows depict the previous work (light black, Section \ref{['sec:SHMARP']}), the flare and SEP coupling process (black, Section \ref{['sec:Matching']}) and the SMARP-SHARP data selection (blue, Section \ref{['sec:Statistics']}).
• Figure 3: Timelines of the total line-of-sight unsigned flux (USFLUXL in Maxwells) and the unsigned flux R near polarity inversion lines (R_VALUE in Maxwells) for the total tracking period of AR$\,$10180 and the day (11/09/2002) during which a SEP occurred (bottom). Underlined (solid green) are the start time (t_start keyword) of the SEP producing (positive) flare of this AR and the data points that the matching algorithm selects for the ML-ready datasets. The process of selecting the data points in the green boxes is referred to as SMARP-SHARP Data Selection in Figure \ref{['fig:Schematic']}.
• Figure 4: Histograms of the minimum prediction windows variation ($t_p \geq t_{start} - 10$ hours) for the 10-hour window ML-ready dataset. The inset histograms showcase the variation of prediction windows within the 1st bin. The data were split into 10 bins while the time between orange and blue dashed lines is equal to the prediction windows' ($t_p$) standard deviation (StD). The histograms in red are denoted the negative event values, and in green are the SEP-producing events.
• Figure 5: Probability density histograms for the six SHARP-SMARP predictors (Table \ref{['tab:SHMARPKeywords']}) for the 10-hour window ML-ready dataset for the positive (red) and negative (green) events. The red and green curves correspond to the fitted trend lines of the histogram bin values.