Skip to main content
NASA Advanced Supercomputing (NAS) Division World-class modeling, simulation, and supercomputing in support of the NASA mission

Intelligent Databases and Machine-Learning Forecast of Solar Flares

AMS Seminar Series

Speaker: Viacheslav (Slava) Sadykov, New Jersey Institute of Technology

Presentation

Seminar Slide Deck (PDF-38.9MB)

Abstract

The growing needs for the accurate Space Weather Forecasts motivate the researchers to implement comprehensive prediction algorithms for solar transient events (flares, coronal mass ejections, filament eruptions). However, each prediction attempt includes the data collection and processing phase, which usually takes a significant part of the research time. Besides some attempts (e.g., SHARP archive of Solar Dynamics Observatory), currently there are no databases containing processed descriptors of the Active Regions (AR) and related observations of flares and eruptive events. In the first part of the talk I will present an Interactive Multi-Instrument Database of Solar Flares (IMIDSF, implemented at https://helioportal.nas.nasa.gov) which is an attempt to integrate the flare records from different satellites and ground-based observatories and provide a search for uniquely-identified flare events based on their physical descriptors. I will discuss the expansion of IMIDSF to the Intelligent Database of Solar Events and Active Regions which also includes the characteristics of the parental ARs and related observations, operational data from Space Weather Prediction Center NOAA, and some new physical data products (Non-Linear Force-Free Field extrapolations and local helioseismology subsurface flow maps) linked to the flares. In the second part of the talk I will give a brief overview of the current attempts to predict the solar flares using the characteristics of the magnetic fields obtained by the HMI instrument on Solar Dynamics Observatory. In this work, the Support Vector Machine (SVM) classifier was trained for the flare prediction of ≥M1.0 and ≥X1.0 flares. Our results indicate that magnetic field Polarity Inversion Line (PIL) characteristics are of special importance for the operational flare forecasts. The obtained True Skill Statistics (TSS) values of 0.76 ± 0.03 for ≥M1.0 flares and 0.84 ± 0.07 for ≥X1.0 flares are better than current expert-based predictions.

Biography

Viacheslav Sadykov is a Ph.D. student (Applied Physics) at the New Jersey Institute of Technology and a Research Associate at the Bay Area Environmental Research Institute. He obtained his master’s degree in Applied Mathematics and Physics in 2015 from the Moscow Institute of Physics and Technology (MIPT). His current research interests include analysis of multiline spectral observations obtained from the NASA's IRIS satellite and modeling of solar spectral lines (radiation transfer problem) observed by IRIS and HMI/SDO performed under the NASA NESSF Fellowship grant “Analysis of Chromospheric Evaporation from IRIS Observations”. His interests also include the development and maintenance of databases of solar flares, and problems of prediction of solar flares and flare-related events. Viacheslav is one of the main contributors to the development of the Interactive Multi-Instrument Database of Solar Flares (IMIDSF, https://helioportal.nas.nasa.gov).