Accurate and reliable predictions of solar flares are essential due to their potentially significant impact on Earth and space-based infrastructure. Although deep learning models have shown notable predictive capabilities in this domain, current evaluations often focus on accuracy while neglecting interpretability and reliability—factors that are especially critical in operational settings. To address this gap, we propose a novel proximity-based framework for analyzing post hoc explanations to assess the interpretability of deep learning models for solar flare prediction. Our study compares two models trained on full-disk line-of-sight (LoS) magnetogram images to predict ≥M-class solar flares within a 24-hour window. We employ the Guided Gradient-weighted Class Activation Mapping (Guided Grad-CAM) method to generate attribution maps from these models, which we then analyze to gain insights into their decision-making processes. To support the evaluation of explanations in operational systems, we introduce a proximity-based metric that quantitatively assesses the accuracy and relevance of local explanations when regions of interest are known. Our findings indicate that the models’ predictions align with active region characteristics to varying degrees, offering valuable insights into their behavior. This framework enhances the evaluation of model interpretability in solar flare forecasting and supports the development of more transparent and reliable operational systems.

Several natural phenomena, such as floods, earth-quakes, volcanic eruptions, or extreme space weather events often come with severity indexes. While these indexes, whether linear or logarithmic are vital, data-driven predictive models for these events rather use a fixed threshold. In this paper, we explore encoding this ordinality to enhance the performance of data-driven models, with specific application in solar flare forecasting. The prediction of solar flares is commonly approached as a binary forecasting problem, categorizing events as either Flare (FL) or No-Flare (NF) based on a chosen threshold (e.g., >C-class, > M-class, or >X-class). However, this binary formulation overlooks the inherent ordinality between the sub-classes within each binary class (FL and NF). In this paper, we propose a novel loss function aimed at optimizing the binary flare prediction problem by embedding the intrinsic ordinal flare characteristics into the binary cross-entropy (BCE) loss function. This modification is intended to provide the model with better guidance based on the ordinal characteristics of the data and improve the overall performance of the models. For our experiments, we employ a ResNet34-based model with transfer learning to predict 2:M-class flares by utilizing the shape-based features of magnetograms of active region (AR) patches spanning from -90° to +90°of solar longitude as our input data. We use a composite skill score (CSS) as our evaluation metric, which is calculated as the geometric mean of the True Skill Score (TSS) and the Heidke Skill Score (HSS) to rank and compare our models' performance. The primary contributions of this work are as follows: (i) We introduce a novel approach to encode ordinality into a binary loss function showing an application to solar flare prediction, (ii) We enhance solar flare forecasting by enabling flare predictions for each AR across the entire solar disk, without any longitudinal restrictions, and evaluate and compare performance. (iii) Our candidate model, optimized with the proposed loss function, shows an improvement of (~17%, (~14%, and (~13% for AR patches within ±30°, ±60°, and ±90° of solar longitude, respectively in terms of CSS, when compared with standard BCE. Additionally, we demonstrate the ability to issue flare forecasts for ARs in near-limb regions (regions between ±60° to ±90°) with a CSS=0.34 (TSS=0.50 and HSS=0.23), expanding the scope of AR-based models for solar flare prediction. This advances the reliability of solar flare forecasts, leading to more effective prediction capabilities.

Forecasting the relevant characteristics of central space weather events such as coronal mass ejections and solar flares is of utmost interest due to their potential near-Earth impact on our technological infrastructure. A key solar feature that is successfully utilized for predicting these solar events is magnetic polarity inversion lines (MPILs). Derived from magnetic field rasters, MPILs represent the shear layers between two opposing (i.e., positive and negative) polarity regions, and the complexity of these separating lines is shown to be highly relevant precursors for these solar events. In this paper, we present our data ecosystem for detecting and serving metadata for MPILs, along with important spatial, temporal and spatial-temporal search capabilities. The MPILs are detected using our detection framework and are served through our public APIs. Our APIs provide active region-based, spatial, and temporal search capabilities. The MPIL metadata consists of a series of binary rasters and metadata parameters. The rasters show the polarity inversion lines, regions of polarity inversion, the unsigned negative and positive polarity regions (both positive and negative) and convex hull of polarity inversion lines, while metadata features include physical and shape-based image parameters. We organize the metadata series as spatial and temporal time series derived from active region trajectories. This data resource is heterogeneous in nature and is designed to be easily extensible. MPIL-derived features are currently used in various deployed operational systems. We envision that our near-real time detection module and API service will be used as the backbone for operational space weather forecasting tools.

Recently, the synergy of physics-based feature engineering and data-intensive methods, including machine learning and deep learning, has ushered in a new era in the analysis and prediction of space weather forecasting, specifically for solar flare prediction. These sophisticated approaches play a pivotal role in understanding the complex mechanisms leading to solar flares, with a primary focus on forecasting these events and mitigating potential risks they pose to our planet. While current methodologies have made substantial advancements, they are not without limitations, and one particularly glaring limitation is the neglect of temporal evolution characteristics within the active regions from which solar flares originate. This oversight impairs the capacity of these methods to capture the intricate relationships among high-dimensional features of these active regions, thereby constraining their practical utility. Our study focuses on two key objectives: the development of interpretable classifiers for multivariate time series data and the introduction of an innovative feature ranking method using sliding window-based sub-interval ranking. The central contribution of our work lies in bridging the gap between complex, less interpretable “black-box” models typically employed for high-dimensional data and the exploration of pertinent sub-intervals within multivariate time series data, with a specific emphasis on solar flare forecasting. Our findings underscore the efficacy of our sliding-window time series forest classifier in solar flare prediction, achieving a True Skill Statistic of over 85%. Our approach is capable of pinpointing the most critical features and sub-intervals relevant to any given learning task. These results indicate significant progress toward improving the interpretability and accuracy of flare prediction models, further advancing our understanding of these impactful events.

This study progresses solar flare prediction research by presenting a full-disk deep-learning model to forecast $\geq\mathrm{M}$-class solar flares and evaluating its efficacy on both central (within ±70°) and near-limb beyond ±70°) events, showcasing qualitative assessment of post hoc explanations for the model’s predictions, and providing empirical findings fro human-centered quantitative assessments of these explanations. Our model is trained using hourly full-disk line-of-sight magnetogram images to predict $\geq{\mathrm{M}}$-class solar flares within the subsequent 24-hour prediction window. Additionally, we apply the Guided Gradient-weighted Class Activation Mapping (Guided Grad-CAM) attribution method to interpret our model’s predictions and evaluate the explanations. Our analysis unveils that full-disk solar flare predictions correspond with active region characteristics. The following points represent the most important findings of our study: ❨1❩ Our deep learning models achieved an average true skill statistic (TSS) of $\sim 0.51$ and a Heidke skill score (HSS) of $\sim.38$, exhibiting skill to predict solar flares where for central locations the average recall is $\sim 0.75$ (recall values for X- and M-class are 0.95 and 0.73 respectively) and for the near-limb flares the average recall is $\sim 0.52$ (recall values for X- and M- class are 0.74 and 0.50 respectively); ❨2❩ qualitative examination of the model’s explanations reveals that it discerns and leverages features linked to active regions in both central and near-limb locations within full-disk magnetograms to produce respective predictions. In essence, our models grasp the shape and texture-based properties of flaring active regions, even in proximity to limb areas—a novel and essential capability with considerable significance for operational forecasting systems.

Prediction of solar flares is a challenging problem in space weather forecasting that has piqued the interest of many researchers in recent years due to improved data availability and the advancements in the field of machine learning and deep learning. In this paper, we present a solution to full-disk flare prediction using compressed magnetogram images, which was performed by training a set of Convolutional Neural Networks to perform operations-ready flare forecasts. W e s elected two prediction modes, which are both binary for predicting the occurrence of ≥M1.0 and ≥C1.0 class flares within the next 24 hours. For this, we use a simple yet powerful pre-trained AlexNet model and we collect compressed images derived from solar magnetograms provided by the Helioseismic and Magnetic Imager (HMI) instrument onboard Solar Dynamics Observatory (SDO). We followed two time-segmented cross-validation strategies: chronological and non-chronological, to effectively understand the predictive skill of our models. We also trained our models using data-augmentation and oversampling to address the existing class-imbalance issue and used true skill statistic (TSS) and Heidke skill score (HSS) as metrics to compare and evaluate. The major results of this study are (1) we successfully implemented an efficient and effective full-disk flare predictor ready for operational forecasting using 8-bit compressed images of solar magnetograms without further preprocessing; (2) Our candidate model achieves an average TSS of 0.47±0.06 for ≥M1.0 mode and 0.63±0.05 for ≥C1.0 mode, and HSS of 0.35±0.05 for ≥M1.0 and 0.62±0.05 for ≥C1.0 mode. Our experimental evaluation also suggests that training a flare prediction model is heavily influenced by the sampling strategies involved due to the imbalanced nature of the datasets and predicting ≥M1.0 class flares is a more challenging task compared to ≥C1.0 ones.

Magnetic polarity inversion line (PIL) in solar active regions have been recognized as essential features for the occurrence of solar flares and the prediction of the flaring phenomenon. In this work, we provide a software framework that detects PILs from the line-of-sight (LoS) or the radial component of the magnetic field vector in active region magnetogram patches. The PIL detection procedure is based on an edge detection technique along with magnetic field strength and PIL size filter. First, we identify positive and negative polarity regions with a magnetic field strength threshold. Then, we utilize the Canny edge detector and morphological operations to both positive and negative regions to identify coarse PILs. Finally, we generate PILs by applying magnetic field strength and PIL size filter to the coarse PILs as mentioned above. Moreover, we provide feature extraction functions to obtain the properties of PILs (i.e., PIL size, the area of polarity inversion, the masked unsigned flux of enclosing PIL, convexity, eigenvalues, fractal dimension, and Hu moments of PIL shape), and produce three PIL-related binary masks (i.e., PIL, the region of polarity inversion, and the convex hull of PIL) for each Longitudinal magnetogram patch. We also provide a qualitative evaluation of our module. The detection results for a well known active region (NOAA AR 11158) are shown as proof of concept.

Outlier detection has become one of the core tasks in spatio-temporal data mining. It plays an essential role in data quality improvement for the machine learning models and recognizing the anomalous patterns, which may remarkably deviate from expected patterns among the trajectory datasets. In this work, we propose a clustering-based technique to detect local outliers in trajectory datasets by utilizing spatial and temporal attributes of moving objects. This local outlier detection involves three phases. In the first phase, we apply a temporal partition procedure to divide the raw trajectory into multiple trajectory segments and extract trajectory features from spatial and temporal attributes for each trajectory segment. Then, we generate template features of trajectory segments by applying a clustering schema in the second phase. Finally, we use the abnormal score - a novel dissimilarity measure, which quantifies the disparity among the query and template trajectory segments in terms of trajectory features and hence determines the local outliers based on the distribution of abnormal score. To demonstrate the effectiveness of our method, we conduct three case studies on the real-life spatio-temporal trajectory datasets from the solar astroinformatics domain (i.e., solar active regions, coronal mass ejections, polarity inversion lines (PIL)). Our experimental results show that our local outlier detection approach can effectively discover the erroneous reports from the reporting module and abnormal phenomenon in various spatio-temporal trajectory datasets.

Image super-resolution is a branch of image processing that is concerned with enhancing the spatial resolution and quality of images by learning the intrinsic details and relations between the lower resolution input and the higher resolution output images. It is widely accepted as an ill-posed problem, which has seen tremendous advancements with deep learning-based models. In this work, we present two super resolution models, Sub-Pixel Convolutional Neural Network (CNN) and Enhanced Deep Residual Networks (ResNet), which can be used for improving the spatial resolution of solar magnetograms. While the ill-posed nature of problem is still a challenge, there are several application areas, including space weather prediction, which can greatly benefit from the improved spatial resolution of solar magnetograms. Along with classical raster inputs we try to improve the model objective by giving HMI Active Region Patches. We show that through our experimental evaluation our models perform better than baselines and CNN-based super resolution model provides viable results for magnetogram super resolution.

Image super-resolution is a branch of image processing that is concerned with enhancing the spatial resolution and quality of images by learning the intrinsic details and relations between the lower resolution input and the higher resolution output images. It is widely accepted as an ill-posed problem, which has seen tremendous advancements with deep learning based models. In this work, we present two magnetogram super resolution models, Sub-Pixel Convolutional Neural Network (CNN) and Enhanced Deep Residual Networks (ResNet), which can be used for improving the spatial resolution of solar magnetograms. While the ill-posed nature of problem is still a challenge, there are several application areas, including space weather prediction, which can greatly benefit from the improved spatial resolution of solar magnetograms. We show that through our experimental evaluation our models perform better than baselines and Sub-Pixel CNN super resolution model provides viable results for magnetogram super resolution.

An all-clear flare prediction is a type of solar flare forecasting that puts more emphasis on predicting non-flaring instances (often relatively small flares and flare quiet regions) with high precision while still maintaining valuable predictive results. While many flare prediction studies do not address this problem directly, all-clear predictions can be useful in operational context. However, in all-clear predictions, finding the right balance between avoiding false negatives (misses) and reducing the false positives (false alarms) is often challenging. Our study focuses on training and testing a set of interval-based time series named Time Series Forest (TSF). These classifiers will be used towards building an all-clear flare prediction system by utilizing multivariate time series data. Throughout this paper, we demonstrate our data collection, predictive model building and evaluation processes, and compare our time series classification models with baselines using our benchmark datasets. Our results show that time series classifiers provide better forecasting results in terms of skill scores, precision and recall metrics, and they can be further improved for more precise all-clear forecasts by tuning model hyperparameters.

In analyses of rare-events, regardless of the domain of application, class-imbalance issue is intrinsic. Although the challenges are known to data experts, their explicit impact on the analytic and the decisions made based on the findings are often overlooked. This is in particular prevalent in interdisciplinary research where the theoretical aspects are sometimes overshadowed by the challenges of the application. To show-case these undesirable impacts, we conduct a series of experiments on a recently created benchmark data, named Space Weather ANalytics for Solar Flares (SWAN-SF). This is a multivariate time series dataset of magnetic parameters of active regions. As a remedy for the imbalance issue, we study the impact of data manipulation (undersampling and oversampling) and model manipulation (using class weights). Furthermore, we bring to focus the auto-correlation of time series that is inherited from the use of sliding window for monitoring flares' history. Temporal coherence, as we call this phenomenon, invalidates the randomness assumption, thus impacting all sampling practices including different cross-validation techniques. We illustrate how failing to notice this concept could give an artificial boost in the forecast performance and result in misleading findings. Throughout this study we utilized Support Vector Machine as a classifier, and True Skill Statistics as a verification metric for comparison of experiments. We conclude our work by specifying the correct practice in each case, and we hope that this study could benefit researchers in other domains where time series of rare events are of interest.

Spatio-temporal co-occurrence analysis captures the spatial and temporal relations between events that occur at the same time and location. In this paper, we utilize spatio-temporal co-occurrence relations to integrate solar active region (AR) data detected and reported by three feature recognition methods, namely, human labeling by forecasters at National Oceanic and Atmospheric Administration (NOAA), Spaceweather HMI Active Region Patch (SHARP) detection pipeline, and Spatial Possibilistic Clustering Algorithm (SPoCA). We determine the associations between individual reports by identifying the spatio-temporal co-occurrences among the reports from these modules. We compare our findings with the data from the Joint Science Operations Center (JSOC), analyzing the discrepancies in different circumstances. We found 105 SHARP series not properly associated with the NOAA-labeled ARs. In the end, we provide detailed movement analyses for the AR trajectories, create an updated SHARP-to-NOAA AR associations, that is crucial for space weather predictions utilizing magnetic field information, and make the ternary associations between SHARP, NOAA ARs, and SPoCA ARs available to the public.

Machine learning-based space weather analytics has attracted much attention due to the potential damages that can be caused by the extreme space weather events. Using a recently released data benchmark, named SWAN-SF, designed for solar flare forecasting based on the pre-flare time series of solar magnetic field parameters, we conduct a case study on the impacts of statistical features derived from the multivariate time series. We investigate the relationship between the number of needed statistical features extracted from the multi-variate time series and the performance of flare forecast models. To that end, we employ random forest and mean decrease impurity to determine a feature selection methodology along with an evaluation procedure. The proposed evaluation method delivers a balance between the two frequently used metrics in this domain, namely True Skill Statistic and Heidke Skill Score. Our approach allows to introduce a generic feature selection and evaluation procedure that is independent from the minor and often obscured decisions that must be made for having a binary forecast model, while presenting interpretable and actionable tools that can help non-data experts make more informed and realistic decisions.

Magnetic parameters are crucial to analyze and forecast solar events such as solar flares and coronal mass ejections. These parameters can be computed from vector magnetogram data collected from the solar surface. We propose a tool to compute these magnetic parameters from solar image data files in parallel using multi-threading and GPUs. The architecture of the proposed solar magnetic parameter computational tool is discussed in detail. We use the images from Spaceweather HMI Active Region Patches (SHARP) data available from JSOC to perform our experiments. We perform exhaustive analysis on the parameters used in the architecture. Run times of magnetic parameter generation are compared across serial execution codes (implemented in Python and C++) and parallel code (implemented in python and C++ using OpenACC). We conclude by showing that parallel computation of magnetic parameters using GPUs is much faster when compared to serial execution.

Solar energetic particles are a result of intense solar events such as solar flares and Coronal Mass Ejections (CMEs). These latter events all together can cause major disruptions to spacecraft that are in Earth's orbit and outside of the magnetosphere. In this work we are interested in establishing the necessary conditions for a major geo-effective solar particle storm immediately after a major flare, namely the existence of a direct magnetic connection. To our knowledge, this is the first work that explores not only the correlations of GOES X-ray and proton channels, but also the correlations that happen across all the proton channels. We found that proton channels autocorrelations and cross-correlations may also be precursors to the occurrence of an SEP event. In this paper, we tackle the problem of predicting >100 MeV SEP events from a multivariate time series perspective using easily interpretable decision tree models.

The automated detection of solar phenomena from high resolution images became important for solar physics researchers after the launch of the Solar Dynamics Observatory. The solar event detections help researchers find and track relevant regions and eventually facilitate the discovery of trends and patterns between different types of events. We address the problem of automated detection of solar events from multi-wavelength solar images using deep learning-based Faster R-CNN method. Earlier work on solar event detection primarily use the observed models to locate the events on the solar images in an unsupervised fashion and each detection algorithm targets specific solar event type. Here, we will present a data-driven methodology to facilitate solar physics research. While this work presents a proof of concept that supervised deep learning-based event detection methodology for solar images is possible, our results show that data-driven detection using deep learning can successfully detect the multiple types of solar events and it can be used for validating the results from existing modules or training modules to detect new event types.

With the advancements in spatiotemporal co-occurrence pattern and event sequence mining algorithms, spatiotemporal knowledge discovery from solar event datasets has been prominent in solar data mining. This work presents an efficient and extensible data access mechanism specifically designed for spatiotemporal relationships among the solar event instances. Previous indexing strategies primarily focus on indexing the trajectories of solar event instances. We propose a graph-based indexing structure for spatiotemporal relationships appearing among the event instances. In our graph structure, the vertices correspond to trajectory-based solar event instances, and the edges represent a particular spatiotemporal relationship. Furthermore, it forms the groundwork for semantic representations of solar event instances.

This paper introduces three interpolation methods that enrich complex evolving region trajectories that are captured every day from numerous ground-based and space-based solar observatories. The interpolation module takes a trajectory as its input and generates an enriched trajectory with interpolated time-geometry pairs. we created three different interpolation techniques that are: MBR-Interpolation (Minimum Bounding Rectangle Interpolation), CP-Interpolation (Complex Polygon Interpolation), and FP-Interpolation (Filament Polygon Interpolation). The methods combine K-means clustering algorithm, shape signature representation, and linear interpolation to generate the missing polygons. This is the first research of this kind that attempts to address the problem of solar big data interpolation. Finally, we outline future improvements and opportunities for solar data interpolation.

One of the main strengths of Geographical Information Systems (GIS) is the analysis of spatial and attributive data. Spatiotemporal interpolation techniques allow the expansion of the collected data to the sites where no samples are available. In the context of GIS, the data, be it interpolated or collected, are visual in nature and hard to understand in raw forms. Visualization of complex evolving region trajectories is often times used as an aid to better understand the data and its underlying patterns. In this work, we created SOLEV, a solar event video generation framework that integrates multiple data sources of solar images. This is the first framework of this kind that not only visualizes spatial solar event boundaries, but also the tracked and interpolated spatiotemporal trajectories they form over time.

Mining spatiotemporal co-occurrence patterns requires assessing the strength of co-occurrences among the instances of different feature types. Currently, a spatiotemporal version of the Jaccard measure is used for measuring the strength of spatiotemporal co-occurrences. We present an extended spatiotemporal version of the Jaccard measure (J*) that is more relevant and efficient for the task of STCOP mining. We also demonstrate the space and time efficiency of the J* with experimental evaluation.

In this paper, we investigate using specifically-designated spatiotemporal indexing techniques for mining cooccurrence patterns from spatiotemporal datasets with evolving polygon-based representations. Previously, suggested techniques for spatiotemporal pattern mining algorithms did not take spatiotemporal indexing techniques into account. We present a new framework for mining spatiotemporal co-occurrence patterns that can use various indexing techniques for efficiently accessing data. Two well-studied spatiotemporal indexing structures, Scalable and Efficient Trajectory Index (SETI) and Chebyshev Polynomial Indexing are currently implemented and available in our framework.

Spatiotemporal co-occurrence patterns (STCOPs) represent the subsets of event types that occur together in both space and time. However, the discovery of STCOPs in data sets with extended spatial representations that evolve over time is computationally expensive because of the necessity to calculate interest measures to assess the co-occurrence strength, and the number of candidates for STCOPs growing exponentially with the number of spatiotemporal event types. In this paper, we introduce a novel and effective filter-and-refine algorithm to efficiently find prevalent STCOPs in massive spatiotemporal data repositories with polygon shapes that move and evolve over time. We provide theoretical analysis of our approach, and follow this investigation with a practical evaluation of our algorithm effectiveness on three real-life data sets and one artificial data set.