Investigating Performance Trends of Simulated Real-time Solar Flare Predictions: The Impacts of Training Windows, Data Volumes, and the Solar Cycle
The Astrophysical Journal, 964:163 (17pp), 2024 April 1
Abstract
This study explores the behavior of machine-learning-based flare forecasting models deployed in a simulated operational environment. Using Georgia State University's Space Weather Analytics for Solar Flares benchmark data set, we examine the impacts of training methodology and the solar cycle on decision tree, support vector machine, and multilayer perceptron performance. We implement our classifiers using three temporal training windows: stationary, rolling, and expanding. The stationary window trains models using a single set of data available before the first forecasting instance, which remains constant throughout the solar cycle. The rolling window trains models using data from a constant time interval before the forecasting instance, which moves with the solar cycle. Finally, the expanding window trains models using all available data before the forecasting instance. For each window, a number of input features (1, 5, 10, 25, 50, and 120) and temporal sizes (5, 8, 11, 14, 17, and 20 months) were tested. To our surprise, we found that, for a window of 20 months, skill scores were comparable regardless of the window type, feature count, and classifier selected. Furthermore, reducing the size of this window only marginally decreased stationary and rolling window performance. This implies that, given enough data, a stationary window can be chosen over other window types, eliminating the need for model retraining. Finally, a moderately strong positive correlation was found to exist between a model's false-positive rate and the solar X-ray background flux. This suggests that the solar cycle phase has a considerable influence on forecasting.
Image Synthesis for Solar Flare Prediction
The Astrophysical Journal Supplement Series, 271:29 (11pp), 2024 March
Image Synthesis for Solar Flare Prediction (iop.org)
Abstract
Solar flare prediction is a topic of interest to many researchers owing to the potential of solar flares to affect various technological systems, both terrestrial and in orbit. In recent years, the forecasting task has become progressively more reliant on data-driven computations and machine-learning algorithms. Although these efforts have improved solar flare predictions, they still falter in doing so for large solar flares, in particular under operational conditions, since large-flare data are very scarce and labeled data are heavily imbalanced. In this work, we seek to address this fundamental issue and present a scheme for generating synthetic magnetograms to reduce the imbalance in the data. Our method consists of (1) synthetic oversampling of line-of-sight magnetograms using Gaussian mixture model representation, followed by (2) a global optimization technique to ensure consistency of both physical features and flare precursors, and (3) the mapping of the generated representations to realistic magnetogram images using deep generative models. We show that these synthetically generated data indeed improve the capacity of solar flare prediction models and that, when tested on such a state-of-the-art model, it significantly enhances its forecasting performance, achieving an F1-score as high as 0.43 ± 0.08 and a true skill statistic of 0.64 ± 0.10 for X-class flares in the 24 hr operational solar flare data split.
Class-Based Time Series Data Augmentation to Mitigate Extreme Class Imbalance for Solar Flare Prediction
Abstract
Time series data plays a crucial role across various domains, making it valuable for decision-making and predictive modeling. Machine learning (ML) and deep learning (DL) have shown promise in this regard, yet their performance hinges on data quality and quantity, often constrained by data scarcity and class imbalance, particularly for rare events like solar flares. Data augmentation techniques offer a potential solution to address these challenges, yet their effectiveness on multivariate time series datasets remains underexplored. In this study, we propose a novel data augmentation method for time series data named Mean Gaussian Noise (MGN). We investigate the performance of MGN compared to eight existing basic data augmentation methods on a multivariate time series dataset for solar flare prediction, SWAN-SF, using a ML algorithm for time series data, TimeSeriesSVC. The results demonstrate the efficacy of MGN and highlight its potential for improving classification performance in scenarios with extremely imbalanced data. Our time complexity analysis shows that MGN also has a competitive computational cost compared to the investigated alternative methods.
Advancing Solar Flare Prediction using Deep Learning with Active Region Patches
Abstract
Solar flares are one of the key space weather phenomena characterized by sudden and intense emissions of radiation from the Sun. The precise and reliable prediction of these phenomena is important due to their potential adverse effects on both space and Earth-based infrastructure. In this paper, we introduce a novel methodology for leveraging shape-based characteristics of magnetograms of active region (AR) patches and provide a novel capability for predicting solar flares covering the entirety of the solar disk (AR patches spanning from -90 ◦ to +90 ◦ of solar longitude). We create three deep learning models: (i) ResNet34, (ii) MobileNet, and (iii) MobileViT to predict ≥M-class flares and assess the efficacy of these models across various ranges of solar longitude. Given the inherent imbalance in our data, we employ augmentation techniques alongside undersampling during the model training phase, while maintaining imbalanced partitions in the testing data for realistic evaluation. We use a composite skill score (CSS) as our evaluation metric, computed as the geometric mean of the True Skill Score (TSS) and the Heidke Skill Score (HSS) to rank and compare models. The primary contributions of this work are as follows: (i) We introduce a novel capability in solar flare prediction that allows predicting flares for each ARs throughout the solar disk and evaluate and compare the performance, (ii) Our candidate model (MobileNet) achieves a CSS=0.51 (TSS=0.60 and HSS=0.44), CSS=0.51 (TSS=0.59 and HSS=0.44), and CSS=0.48 (TSS=0.56 and HSS=0.40) for AR patches within ±30 ◦ , ±60 ◦ , ±90 ◦ of solar longitude respectively. Additionally, we demonstrate the ability to issue flare forecasts for ARs in near-limb regions (regions between ±60 ◦ to ±90 ◦ ) with a CSS=0.39 (TSS=0.48 and HSS=0.32), expanding the scope of AR-based models for solar flare prediction. This advancement opens new avenues for more reliable prediction of solar flares, thereby contributing to improved forecasting capabilities.
A novel solar flare forecast model with deep convolution neural network and one-against-rest approach
Advances in Space Research Available online 19 June 2024
Abstract
We present a novel deep Convolutional Neural Network model with one-against-rest approach (OAR-CNN) and modify the hybrid Convolutional Neural Network (H-CNN) model of Zheng et al. (2019) for multiclass flare prediction to forecast whether an active region generates multiclass flare within 24 h. Additionally, in the OAR-CNN and H-CNN models, we employ the decision strategies of majority voting and probability threshold, respectively, comparing the prediction outcomes of these two strategies. Our models undergo training and testing on the same 10 cross-validation datasets as employed by Zheng et al. (2019), and then compare the results with previous studies using forecast verification metrics, with a focus on the true skill statistic (TSS). The major results are summarized as follows. (1) This is the first attempt to utilize the decision strategies of majority voting and probability threshold in the OAR-CNN model for multiclass solar flare prediction. (2) In both the OAR-CNN and H-CNN models, the predictive results with the probability threshold decision strategy are higher than those with majority voting across all six classes (i.e., No-flare, C, M, X, ⩾C, and ⩾M class), except for a slight decrease in the C class in the OAR-CNN model. (3) The OAR-CNN and modified H-CNN models with the probability threshold decision strategy demonstrate comparable statistical scores across all categories and outperform previous studies. (4) In the prediction of four-class flare, our proposed OAR-CNN model with the probability threshold decision strategy achieves relatively high mean TSS scores of 0.744, 0.429, 0.567, and 0.630 for No-flare, C, M, and X class, respectively, surpassing or comparable to results from prior studies. Furthermore, our model achieves high TSS scores of 0.744 ± 0.042 for ⩾C--class and 0.764 ± 0.089 for ⩾M-class predictions.
Anticipating Solar Flares
Abstract
Solar flares commonly have a "hot onset precursor event" (HOPE), detectable from soft X-ray observations. This requires subtraction of pre-flare fluxes from the non-flaring Sun prior to the event, fitting an isothermal emission model to the flare excess fluxes by comparing the GOES passbands at 1--8 ˚A and 0.5--4 ˚A, and plotting the timewise evolution of the flare emission in a diagram of temperature vs emission measure. The HOPE then appears as an initial "horizontal branch" in this diagram. It precedes the non-thermal impulsive phase of the flare and thus the flare peak in soft X-rays as well. We use this property to define a "flare anticipation index" (FAI), which can serve as an alert for observational programs aimed at solar flares based on near-real-time soft X-ray observations. This FAI gives lead times of a few minutes and produces very few false positive alerts even for flare brightenings too weak to merit NOAA classification.
A Deep Learning Approach to Operational Flare Forecasting
A Deep Learning Approach to Operational Flare Forecasting (arxiv.org)
Abstract
Solar flares are explosions on the Sun. They happen when energy stored in magnetic fields around solar active regions (ARs) is suddenly released. In this paper, we present a transformerbased framework, named SolarFlareNet, for predicting whether an AR would produce a γ-class flare within the next 24 to 72 hours. We consider three γ classes, namely the ≥M5.0 class, the ≥M class and the ≥C class, and build three transformers separately, each corresponding to a γ class. Each transformer is used to make predictions of its corresponding γ-class flares. The crux of our approach is to model data samples in an AR as time series and to use transformers to capture the temporal dynamics of the data samples. Each data sample consists of magnetic parameters taken from Space-weather HMI Active Region Patches (SHARP) and related data products. We survey flare events that occurred from May 2010 to December 2022 using the Geostationary Operational Environmental Satellite X-ray flare catalogs provided by the National Centers for Environmental Information (NCEI), and build a database of flares with identified ARs in the NCEI flare catalogs. This flare database is used to construct labels of the data samples suitable for machine learning. We further extend the deterministic approach to a calibration-based probabilistic forecasting method. The SolarFlareNet system is fully operational and is capable of making near real-time predictions of solar flares on the Web.