Feature Ranking of Active Region Source Properties in Solar Flare Forecasting and the Uncompromised Stochasticity of Flare Occurrence

Abstract

Solar flares originate from magnetically active regions but not all solar active regions give rise to a flare. Therefore, the challenge of solar flare prediction benefits by an intelligent computational analysis of physics-based properties extracted from active region observables, most commonly line-ofsight or vector magnetograms of the active-region photosphere. For the purpose of flare forecasting, this study utilizes an unprecedented 171 flare-predictive active region properties, mainly inferred by the Helioseismic and Magnetic Imager onboard the Solar Dynamics Observatory (SDO/HMI) in the course of the European Union Horizon 2020 FLARECAST project. Using two different supervised machine learning methods that allow feature ranking as a function of predictive capability, we show that: i) an objective training and testing process is paramount for the performance of every supervised machine learning method; ii) most properties include overlapping information and are therefore highly redundant for flare prediction; iii) solar flare prediction is still -and will likely remain -a predominantly probabilistic challenge. Nishizuka et al 2018) within the recent field of space weather forecasting that relies on the availability of two ingredients; one observational and one computational. First, it is well-established that solar active regions (ARs) exclusively host major flares and therefore flare prediction needs experimental data on AR properties, associated to the photospheric and coronal magnetic field; however, coronal information has only recently started being used in the form of EUV images given as input to a deep learning network by Nishizuka et al (2018). Second, this information on AR magnetic properties can be processed for prediction purposes by means of a computational method for data analysis; machine learning has recently offered strong candidates for such methods.Since February 2010, the Helioseismic and Magnetic Imager onboard the Solar Dynamics Observatory (SDO/HMI) (Scherrer et al 2012) is providing both line-of-sight and vector magnetograms of the full solar disk at a (vector magnetogram) cadence of 12 minutes. SDO/HMI magnetograms can be used for solar flare prediction according to two different approaches. First, HMI magnetograms are utilized to calculate a variety of properties either from the line-of-sight component only, from the radial component only, or from all three vector components. Various single-valued quantities, hereafter referred to as features, can be calculated from these property images through a variety of techniques (e.g., thresholding, feature recognition, etc), such that calculation of one physical property may provide multiple features as inputs to machine learning (i.e., image maximum, total, and moments). Of course, additional features that are not derived from property images may also contribute to the input dataset. Second, from a deep learning perspective, HMI images can be given as input to Convolutional Neural Networks (CNNs) that automatically perform a probabilistic f...

Leveraging the mathematics of shape for solar magnetic eruption prediction

“…Neural networks, which go beyond simple binary classification by learning complex nonlinear relationships among their inputs, have also been used to great advantage by Nishizuka et al (2018). A variety of other ML algorithms, such as Bayesian networks (Yu et al, 2010), radial basis model networks (Colak and Qahwaji, 2009), logistic regression (Yuan et al, 2010), LASSO regression (Campi et al, 2019), and random forests or ERTs (Nishizuka et al, 2017;Campi et al, 2019) have also been implemented for solar flare prediction with some degree of success. Nishizuka et al (2017) and Jonas et al (2018) include Solar Dynamics Observatory (SDO) Atmospheric Imaging Assembly and EUV image characteristics in addition to magnetogram data in a fully connected neural network architecture.…”

Solar Flare Prediction Using Magnetic Field Diagnostics above the Photosphere

“…Most flare and CME forecast methods apply photospheric magnetic and Doppler data of ARs for forecasting. Some recent, pioneering approaches with various degrees of success attempt to incorporate solar atmospheric extreme ultraviolet (EUV) data and/or use machine learning in order to improve forecasting accuracy (see, e.g., Qahwaji & Colak 2007;Bobra & Couvidat 2015;Florios et al 2018;Campi et al 2019;Kim et al 2019;Wang et al 2019). Detailed information on measuring, and the consequent modeling, of the 3D magnetic field structure of an AR would be important to obtain more accurate insight into the preflare evolution locally in the solar atmosphere.…”

Testing and Validating Two Morphological Flare Predictors by Logistic Regression Machine Learning

“…In this manner, there is no overlap between training and testing. To ensure robustness of the results, we replicated 100 times the training and test datasets for 6/12/18 and 24-hr issuing time intervals, like e.g., Campi et al (2019).…”

“…However, further quantities derived from AR observations allow a physical comparison and deeper understanding of the actual causes of the solar eruptions. In this sense, different morphological parameters have been introduced to characterised the magnetic field configuration or highlight the existence of polarity-inversion-lines (PILs) in ARs, with varying sophistication (see e.g., Barnes et al, 2016;Leka et al, 2018;Campi et al, 2019;Leka et al, 2019a, Leka et al, 2019bPark et al, 2020, and references therein). Furthermore, Kontogiannis et al (2018) investigated and tested some of those parameters, which were identified as efficient flare predictors.…”

“…The observed magnetic properties of an AR can be processed for the purpose of prediction by machine learning (ML) computational methods for data analysis (Camporeale, 2019), such as neural networks (Ahmed et al, 2013), support vector machines (Bobra and Couvidat, 2015;Boucheron et al, 2015), relevance vector machines (Al-Ghraibah et al, 2015), ordinal logistic regression (Song et al, 2009), decision trees (Yu et al, 2009), random forests (Liu et al, 2017;Domijan et al, 2019), and deep learning (Nishizuka et al, 2018). Notably, parameters B eff , E Ising , G S , and I NN,tot were used by the FLARECAST project 2 , where the prediction capabilities of almost 200 parameters were tested by the LASSO and Random Forest ML techniques (Campi et al, 2019). From these 200 parameters, the FLARECAST project found that the four morphological parameters were ranked as good flare predictors.…”