Solar Flare Forecasting Research Articles

Causal Attention Deep-learning Model for Solar Flare Forecasting

Solar flares originate from the sudden release of energy stored in the magnetic field of the active region on the Sun, but the trigger for flares is still uncertain. Currently, deep-learning-based solar flare prediction models have achieved good results and are widely recognized. However, these models focus more on data correlation rather than causality. An ideal flare prediction model should probe into the causes/triggers of solar flares, and diagnose the precursors of flares rather than just correlation analysis. To extract more informative precursors of flares from magnetograms, while suppressing the interference of confounding factors, a causal attention module is introduced to disentangle causal and confounder features from the input features. To address the problem of imbalanced positive and negative samples in the data set, an adaptive data set split mechanism is proposed. It divides the data set into several balanced subsets of positive and negative samples, and dynamically adjusts the subsets according to the model’s prediction results during the training process. The experimental results demonstrate that our proposed model achieves 4.08%, 8.38%, and 2.19% higher accuracy, true skill score, and area under the receiver operating characteristic curve than the baseline model. Additionally, the class-specific heatmaps by using the gradient-weighted class activation mapping method reveal that our proposed model generally focuses on the polarity inverse line of active regions, well in line with theoretical study.

Time-Series Feature Selection for Solar Flare Forecasting

Solar flares are significant occurrences in solar physics, impacting space weather and terrestrial technologies. Accurate classification of solar flares is essential for predicting space weather and minimizing potential disruptions to communication, navigation, and power systems. This study addresses the challenge of selecting the most relevant features from multivariate time-series data, specifically focusing on solar flares. We employ methods such as Mutual Information (MI), Minimum Redundancy Maximum Relevance (mRMR), and Euclidean Distance to identify key features for classification. Recognizing the performance variability of different feature selection techniques, we introduce an ensemble approach to compute feature weights. By combining outputs from multiple methods, our ensemble method provides a more comprehensive understanding of the importance of features. Our results show that the ensemble approach significantly improves classification performance, achieving values 0.15 higher in True Skill Statistic (TSS) values compared to individual feature selection methods. Additionally, our method offers valuable insights into the underlying physical processes of solar flares, leading to more effective space weather forecasting and enhanced mitigation strategies for communication, navigation, and power system disruptions.

Time Series of Magnetic Field Parameters of Merged MDI and HMI Space-weather Active Region Patches as Potential Tool for Solar Flare Forecasting

Solar flare prediction studies have been recently conducted with the use of Space-Weather MDI (Michelson Doppler Imager on board Solar and Heliospheric Observatory) Active Region Patches (SMARPs) and Space-Weather HMI (Helioseismic and Magnetic Imager on board Solar Dynamics Observatory) Active Region Patches (SHARPs), which are two currently available data products containing magnetic field characteristics of solar active regions (ARs). The present work is an effort to combine them into one data product, and perform some initial statistical analyses in order to further expand their application in space-weather forecasting. The combined data are derived by filtering, rescaling, and merging the SMARP and SHARP parameters, which can then be spatially reduced to create uniform multivariate time series. The resulting combined MDI–HMI data set currently spans the period between 1996 April 4 and 2022 December 13, and may be extended to a more recent date. This provides an opportunity to correlate and compare it with other space-weather time series, such as the daily solar flare index or the statistical properties of the soft X-ray flux measured by the Geostationary Operational Environmental Satellites. Time-lagged cross correlation indicates that a relationship may exist, where some magnetic field properties of ARs lead the flare index in time. Applying the rolling-window technique makes it possible to see how this leader–follower dynamic varies with time. Preliminary results indicate that areas of high correlation generally correspond to increased flare activity during the peak solar cycle.

Solar Flare Forecasting Based on Magnetogram Sequences Learning with Multiscale Vision Transformers and Data Augmentation Techniques

Solar Flare Forecasting Based on Magnetogram Sequences Learning with Multiscale Vision Transformers and Data Augmentation Techniques

Study of Sample Imbalance in Deep Learning Modeling of Solar Flare Forecasting

Study of Sample Imbalance in Deep Learning Modeling of Solar Flare Forecasting

Toward Model Compression for a Deep Learning–Based Solar Flare Forecast on Satellites

Timely solar flare forecasting is challenged by the delay of transmitting vast amounts of data from the satellite to the ground. To avoid this delay, it is expected that forecasting models will be deployed on satellites. Thus, transmitting forecasting results instead of huge volumes of observation data would greatly save network bandwidth and reduce forecasting delay. However, deep-learning models have a huge number of parameters so they need large memory and strong computing power, which hinders their deployment on satellites with limited memory and computing resources. Therefore, there is a great need to compress forecasting models for efficient deployment on satellites. First, three typical compression methods, namely knowledge distillation, pruning, and quantization, are examined individually for compressing of solar flare forecasting models. And then, an assembled compression model is proposed for better compressing solar flare forecasting models. The experimental results demonstrate that the assembled compression model can compress a pretrained solar flare forecasting model to only 1.67% of its original size while maintaining forecasting accuracy.

Deep neural networks of solar flare forecasting for complex active regions

Solar flare forecasting is one of major components of operational space weather forecasting. Complex active regions (ARs) are the main source producing major flares, but only a few studies are carried out to establish flare forecasting models for these ARs. In this study, four deep learning models, called Complex Active Region Flare Forecasting Model (CARFFM)-1, −2, −3, and −4, are established. They take AR longitudinal magnetic fields, AR vector magnetic fields, AR longitudinal magnetic fields and the total unsigned magnetic flux in the neutral line region, AR vector magnetic fields and the total unsigned magnetic flux in the neutral region as input, respectively. These four models can predict the production of M-class or above flares in the complex ARs for the next 48 h. Through comparing the performance of the models, CARFFM-4 has the best forecasting ability, which has the most abundant input information. It is suggested that more valuable and rich input can improve the model performance.

Research Progress on Solar Flare Forecast Methods Based on Data-driven Models

Eruption of solar flares is a complex nonlinear process, and the rays and high-energy particles generated by such an eruption are detrimental to the reliability of space-based or ground-based systems. So far, there are not reliable physical models to accurately account for the flare outburst mechanism, but a lot of data-driven models have been built to study a solar flare and forecast it. In the paper, the status of solar-flare forecasting is reviewed, with emphasis on the machine learning methods and data-processing techniques used in the models. At first, the essential forecast factors strongly relevant to solar flare outbursts, such as classification information of the sunspots and evolution pattern of the magnetic field, are reviewed and analyzed. Subsequently, methods of resampling for data preprocessing are introduced to solve the problems of class imbalance in the solar flare samples. Afterwards, typical model structures adopted for flare forecasting are reviewed from the aspects of the single and fusion models, and the forecast performances of the different models are analyzed. Finally, we herein summarize the current research on solar flare forecasting and outline its development trends.

Deep Learning–based Solar Flare Forecasting Model. III. Extracting Precursors from EUV Images

A solar flare is one of the most intense solar activities, and flare forecasting is necessary to avoid its destructive impact on the near-Earth space environment and technological infrastructure. Previous studies have demonstrated the importance of the photospheric magnetic field in the occurrence of flares. Therefore, most of the input data in traditional solar flare forecasting models are magnetograms of active regions. The magnetic field of the photosphere is routinely measured and observed, but the magnetic field of the corona is not. Hence, the goal of our work is to test whether precursors can be extracted from coronal multiwavelength images of active regions and to build a flare-forecasting model. Therefore, we investigated the effect of using extreme ultraviolet (EUV) images (at 94, 131, 171, 193, 211, and 335 Å) of the active region on solar flare forecasting. We generated a data set consisting of EUV images of the active regions observed by the Solar Dynamics Observatory/Atmospheric Imaging Assembly from 2010 to 2016. Based on this data set, a deep-learning method was used to extract precursors from EUV multiwavelength images. The test results of the forecasting model were discussed and analyzed, and the following conclusions were drawn. (1) Each wavelength achieved good results using the EUV multiwavelength images for flare forecasting. The 94 Å wavelength demonstrated the best result among the single-wavelength results. (2) Among the combined multiwavelength results, the best fusion results were obtained for all six wavelengths.

A Large-Scale Dataset of Three-Dimensional Solar Magnetic Fields Extrapolated by Nonlinear Force-Free Method

It has been widely accepted that solar magnetic field manipulates all solar activities, especially violent solar bursts in solar corona. Thus, it is extremely important to reconstruct three-dimentional (3D) magnetic field of solar corona from really observed photospheric magnetogram. In this paper, a large-scale dataset of 3D solar magnetic fields of active regions is built by using the nonlinear force-free magnetic field (NLFFF) extrapolation from vector magnetograms of Helioseismic and Magnetic Imager (HMI) on Solar Dynamics Observatory (SDO). In this dataset, all space-weather HMI active region patches (SHARPs) with the corresponding serial numbers of national oceanic and atmospheric administration (NOAA) are included. They are downloaded from the SHARP 720 s series of JSOC every 96 minutes. In addition, each sample is labelled with a finer grained label for solar flare forecast. This paper is with the purpose of open availability of data resource and source code to the peers for refraining from repeated labor of data preparation. Meanwhile, with such a large-scale, high spatio-temporal resolution and high quality scientific data, we anticipate a wide attention and interest from artificial intelligence (AI) and computer vision communities, for exploring AI for astronomy over such a large-scale dataset.

Application of Deep Reinforcement Learning to Major Solar Flare Forecasting

In this study, we present the application of deep reinforcement learning to the forecasting of major solar flares. For this, we consider full-disk magnetograms at 00:00 UT from the Solar and Heliospheric Observatory/Michelson Doppler Imager (1996–2010) and the Solar Dynamics Observatory/Helioseismic and Magnetic Imager (2011–2019), as well as Geostationary Operational Environmental Satellite X-ray flare data. We apply Deep Q-Network (DQN) and Double DQN, which are popular deep reinforcement learning methods, to predict “Yes or No” for daily M- and X-class flare occurrence. The reward functions, consisting of four rewards for true positive, false positive, false negative, and true negative, are used for our models. The major results of this study are as follows. First, our deep-learning models successfully predict major solar flares with good skill scores, such as HSS, F1, TSS, and ApSS. Second, the performance of our models depends on the reward function, learning method, and target agent update time. Third, the performance of our deep-learning models is noticeably better than that of a convolutional neural network (CNN) model with the same structure: 0.38 (CNN) to 0.44 (ours) for HSS, 0.47 to 0.52 for F1, 0.53 to 0.59 for TSS, and 0.09 to 0.12 for ApSS.

Deep Learning Based Solar Flare Forecasting Model. II. Influence of Image Resolution

Due to the accumulation of solar observational data and the development of data-driven algorithms, deep learning methods are widely applied to build a solar flare forecasting model. Most of the works focus on how to design or select proper deep networks for the forecasting task. Nevertheless, the influence of image resolution on the learning based solar flare forecasting model has not been analyzed and discussed. In this Paper, we investigate the influence of the resolution of magnetograms on the accuracy of solar flare forecasting. We study the active regions by the Solar Dynamics Observatory/Helioseismic and Magnetic Imager (SDO/HMI) magnetograms from 2010 to 2019. Then, we downsample them to get a database containing active regions with several resolutions. Afterwards, three deep neural networks (i) AlexNet, (ii) ResNet-18, and (iii) SqueezeNet are implemented to evaluate the performance of solar flare forecasting compared to different resolutions of magnetogram. In experiments, we first did comparative experiments on our own simulated HMI database with different resolutions. Then we conducted experiments on two selected actual overlapping databases, Hinode–HMI and Michelson Doppler Imager–HMI, to reconfirm our conclusions. The experiment results show that all the selected deep learning networks are insensitive to the resolution to a certain extent. We visualized the regions of interest of the network from an interpretable perspective and found that the deep learning network pays more attention to the global features extracted from active regions that are not sensitive to local information in magnetograms.

A Comparative Analysis of Machine-learning Models for Solar Flare Forecasting: Identifying High-performing Active Region Flare Indicators

Solar flares create adverse space weather impacting space- and Earth-based technologies. However, the difficulty of forecasting flares, and by extension severe space weather, is accentuated by the lack of any unique flare trigger or a single physical pathway. Studies indicate that multiple physical properties contribute to active region flare potential, compounding the challenge. Recent developments in machine learning (ML) have enabled analysis of higher-dimensional data leading to increasingly better flare forecasting techniques. However, consensus on high-performing flare predictors remains elusive. In the most comprehensive study to date, we conduct a comparative analysis of four popular ML techniques (k nearest neighbors, logistic regression, random forest classifier, and support vector machine) by training these on magnetic parameters obtained from the Helioseismic and Magnetic Imager on board the Solar Dynamics Observatory for the entirety of solar cycle 24. We demonstrate that the logistic regression and support vector machine algorithms perform extremely well in forecasting active region flaring potential. The logistic regression algorithm returns the highest true skill score of 0.967 ± 0.018, possibly the highest classification performance achieved with any strictly parametric study. From a comparative assessment, we establish that magnetic properties like total current helicity, total vertical current density, total unsigned flux, R_VALUE, and total absolute twist are the top-performing flare indicators. We also introduce and analyze two new performance metrics, namely, severe and clear space weather indicators. Our analysis constrains the most successful ML algorithms and identifies physical parameters that contribute most to active region flare productivity.

Solar Flare Forecast Model Based on Resampling and Fusion Method

Predicting the activity of solar flares is of great significance for studying its physical mechanism and the impact on human production and life. Problems such as class imbalance, high time-series sensitivity, and over-localization of important features exist in the sample data used for flare forecasting. We design a solar flare fusion method based on resampling and the CNN-GRU algorithm to try to solve the above problems. In order to verify the effectiveness of this fusion method, first, we compared the forecast performance of different resampling methods by keeping the forecast model unchanged. Then, we used the resampling algorithm with high performance to combine some single forecast models and fusion forecast models respectively. We use the 2010–2017 sunspot data set to train and test the performance of the flare model in predicting flare events in the next 48 h. Through the conclusion of the above steps, we prove that the resampling method SMOTE and its variant SMOTE-ENN are more advantageous in class imbalance problem of flare samples. In addition, after the fusion of one-dimensional convolution and recurrent network with “forget-gate” , combined with the SMOTE-ENN to achieve TSS = 61%, HSS = 61%, TPRate = 77% and TNRate = 83%. This proves that the fusion model based on resampling and the CNN-GRU algorithm is more suitable for solar flare forecasting.

Artificial intelligence generated solar farside magnetogram using conditional generative adversarial network

A solar flare occurs due to a magnetic field reconnection above the active region. The active region magnetic complexity observed in the magnetogram could be used as proxies for solar flare forecasting. It is also known that solar flares that occur from emerging active regions located near the solar disk eastern limb can still have an impact on the Earth. Therefore, magnetic observation of active regions in the solar farside is important to forecast east limb flares occurrences. This study utilizes the conditional Generative Adversarial Network (cGAN) model to generate Artificial Intelligence (AI) solar farside magnetogram. Our model was trained using the solar frontside observations dataset from Solar Dynamic Observatory (SDO)/Atmospheric Imaging Assembly (AIA) 304 Angstrom as input images and SDO/Helioseismic and Magnetic Imager (HMI) magnetogram as output images. Our model generates solar farside magnetogram using solar farside observation from Solar Terrestrial Relations Observatory (STEREO)/Extreme Ultraviolet Imager (EUVI) 304 Angstrom. We also conducted validation on the similarity of our AI-generated magnetogram with the magnetogram observation from SDO/HMI using the Structural Similarity Index (SSIM) method. SSIM obtained an average similarity value of 0.61±0.06 for training data and 0.47±0.02 for validation data which contain active regions producing flares.

МЕТОД ВІДНОШЕННЯ ЛІНІЙ – МОЖЛИВИЙ ІНСТРУМЕНТ ДЛЯ НАДІЙНІШОГО ПРОГНОЗУВАННЯ СОНЯЧНИХ СПАЛАХІВ

Already published and new observational data are given, which indicate that in the regions of solar flares there is a significant difference in the sub-telescopic structure of the magnetic field, which is absent in non-flare regions. In particular, the magnetic fields by the FeI 6301.5 and FeI 6302.5 Å lines measured in flares outside sunspot umbra have an inverse ratio of Bobs(6301.5) < Bobs(6302.5), while without flares it is mostly Bobs(6301.5) > Bobs(6302.5). The conclusion is argued that in flares this is due to the fact that at the photospheric level there is a spatially unresolved contact of flux tubes with “kilogauss” fields and the weaker background field, which have opposite magnetic polarities. Direct observational indications of particularly strong fields of opposite polarity are found in the far wings of the FeI 6173.3 Å line. New unique observational data relating to the region of the seismic source in the exceptionally powerful solar flare of October 28, 2003 of X17.2/4B class indicate that the above inverse line ratio is most significant precisely in the region of this source. A new idea for more reliable forecasting of solar flares is proposed, which boils down to the fact that new relevant information for forecasting should be obtained from monitoring observational data in the FeI 6301.5 and FeI 6302.5 Å lines.

Fine-grained Solar Flare Forecasting Based on the Hybrid Convolutional Neural Networks*

Abstract Improving the performance of solar flare forecasting is a hot topic in the solar physics research field. Deep learning has been considered a promising approach to perform solar flare forecasting in recent years. We first used the generative adversarial networks (GAN) technique augmenting sample data to balance samples with different flare classes. We then proposed a hybrid convolutional neural network (CNN) model (M) for forecasting flare eruption in a solar cycle. Based on this model, we further investigated the effects of the rising and declining phases for flare forecasting. Two CNN models, i.e., M rp and M dp, were presented to forecast solar flare eruptions in the rising phase and declining phase of solar cycle 24, respectively. A series of testing results proved the following. (1) Sample balance is critical for the stability of the CNN model. The augmented data generated by GAN effectively improved the stability of the forecast model. (2) For C-class, M-class, and X-class flare forecasting using Solar Dynamics Observatory line-of-sight magnetograms, the means of the true skill statistics (TSS) scores of M are 0.646, 0.653, and 0.762, which improved by 20.1%, 22.3%, and 38.0% compared with previous studies. (3) It is valuable to separately model the flare forecasts in the rising and declining phases of a solar cycle. Compared with model M, the means of the TSS scores for No-flare, C-class, M-class, and X-class flare forecasting of the M rp improved by 5.9%, 9.4%, 17.9%, and 13.1%, and those of the M dp improved by 1.5%, 2.6%, 11.5%, and 12.2%.

Studying the conditions for magnetic reconnection in solar flares with and without precursor flares

Forecasting of solar flares remains a challenge due to the limited understanding of the underlying physical processes of these events, in particular magnetic reconnection. Studies have indicated that changes to the photospheric magnetic fields associated with magnetic reconnection – particularly in relation to the field helicity – occur during solar flare events. This study utilized data from the Solar Dynamics Observatory (SDO) Helioseismic and Magnetic Imager (HMI) and SpaceWeather HMI Active Region Patches (SHARPs) to analyze full vector-field component data of the photospheric magnetic field during solar flare events within a near decade long HMI dataset. Analysis of the data was used to identify and compare the trends of differing flare classes for varying time intervals leading up to an event, as well as the trends of flares that occur with and without a precursor flare, in order to discern signatures of the physical mechanisms involved. The data suggests that active regions that produce flares with precursors are continuing to evolve and appreciably more complex than those that produce a flare without a precursor. Additionally, precursor flares were found to enhance the shear across an active region, helping set up the conditions necessary for a larger solar flare to occur. Ultimately none of the SHARP parameters showed a distinct signature of magnetic reconnection.

How to Train Your Flare Prediction Model: Revisiting Robust Sampling of Rare Events

Abstract We present a case study of solar flare forecasting by means of metadata feature time series, by treating it as a prominent class-imbalance and temporally coherent problem. Taking full advantage of pre-flare time series in solar active regions is made possible via the Space Weather Analytics for Solar Flares (SWAN-SF) benchmark data set, a partitioned collection of multivariate time series of active region properties comprising 4075 regions and spanning over 9 yr of the Solar Dynamics Observatory period of operations. We showcase the general concept of temporal coherence triggered by the demand of continuity in time series forecasting and show that lack of proper understanding of this effect may spuriously enhance models’ performance. We further address another well-known challenge in rare-event prediction, namely, the class-imbalance issue. The SWAN-SF is an appropriate data set for this, with a 60:1 imbalance ratio for GOES M- and X-class flares and an 800:1 imbalance ratio for X-class flares against flare-quiet instances. We revisit the main remedies for these challenges and present several experiments to illustrate the exact impact that each of these remedies may have on performance. Moreover, we acknowledge that some basic data manipulation tasks such as data normalization and cross validation may also impact the performance; we discuss these problems as well. In this framework we also review the primary advantages and disadvantages of using true skill statistic and Heidke skill score, two widely used performance verification metrics for the flare-forecasting task. In conclusion, we show and advocate for the benefits of time series versus point-in-time forecasting, provided that the above challenges are measurably and quantitatively addressed.

Visual Explanation of a Deep Learning Solar Flare Forecast Model and Its Relationship to Physical Parameters

Abstract In this study, we present a visual explanation of a deep learning solar flare forecast model and its relationship to physical parameters of solar active regions (ARs). For this, we use full-disk magnetograms at 00:00 UT from the Solar and Heliospheric Observatory/Michelson Doppler Imager and the Solar Dynamics Observatory/Helioseismic and Magnetic Imager, physical parameters from the Space-weather HMI Active Region Patch (SHARP), and Geostationary Operational Environmental Satellite X-ray flare data. Our deep learning flare forecast model based on the Convolutional Neural Network (CNN) predicts “Yes” or “No” for the daily occurrence of C-, M-, and X-class flares. We interpret the model using two CNN attribution methods (guided backpropagation and Gradient-weighted Class Activation Mapping [Grad-CAM]) that provide quantitative information on explaining the model. We find that our deep learning flare forecasting model is intimately related to AR physical properties that have also been distinguished in previous studies as holding significant predictive ability. Major results of this study are as follows. First, we successfully apply our deep learning models to the forecast of daily solar flare occurrence with TSS = 0.65, without any preprocessing to extract features from data. Second, using the attribution methods, we find that the polarity inversion line is an important feature for the deep learning flare forecasting model. Third, the ARs with high Grad-CAM values produce more flares than those with low Grad-CAM values. Fourth, nine SHARP parameters such as total unsigned vertical current, total unsigned current helicity, total unsigned flux, and total photospheric magnetic free energy density are well correlated with Grad-CAM values.