Space Weather Prediction System providing forecasts and alerts on solar flares and SEP events

Abstract

A web-based prototype system for predicting Solar Flares and Solar Energetic Particle (SEP) events for its use by space launcher operators or any interested user has been implemented. The main goal of this system, called SEPsFLAREs, is to provide warnings/predictions with forecast horizons from 48 hours before to a few hours before to the SEP peak flux, and duration predictions. The module responsible for predicting solar flares, the SF_PMod, is based on the well-known ASAP flare predictor [T. Colak & R. Qahwaji, Automated solar activity prediction: A hybrid computer platform using machine learning and solar imaging for automated prediction of solar flares, Space Weather, 7 (S06001), 2009], which learns rules by using machine learning techniques on SDO/SOHO solar images to automatically detect sunspots, classify them based on the McIntosh classification system, and predict C-, M-, and X-class flares with forecast horizon from 6 h to 48 h. Regarding the performance of the flare predictor, the 24-hour forecast horizon was found to provide the best performance: the Probability of Detection (POD), False Alarm Ratio (FAR) and True Skill Statistics estimations were 63.8%, 99.0% and 0.5 respectively for predicting X-class flares; and 88.7%, 87.0% and 0.59 respectively, for predicting M-class flares. The module responsible for predicting the SEP onset and occurrence, the SEP_OO_PMod, is based on the well-known UMASEP predictor [M. Núñez, Predicting solar energetic proton events (E > 10 MeV), Space Weather, 9 (S07003), 2011], which performs X-ray and proton flux correlations to find the first symptoms of future well- and poorly-connected SEP events. The SEP_OO_PMod also provides a Warning Tool which is able to warn about potential proton enhancements (including SEP events) from flare predictions. Regarding the performance of the SEP_OO_PMod, it was validated taking into account all 129 SEP events from January 1994 to June 2014 and obtained a POD of 86.82%, a FAR of 25.83%, and an Average Warning Time (AWT) of 3.93 h. Regarding the evaluation of the Warning Tool, the best performance, obtained with a set of user-defined parameters, were a POD of 58.3%, FAR of 90.1%, and AWT of 23.1 h. The module responsible for predicting SEP peak and duration, the SEP_FID_PMod, identifies the parent solar flare associated to an observed/predicted SEP, simulates the radial propagation of the predicted shock on a representative IMF structure (i.e. a static Parker Spiral), and predicts the SEP peak and duration. The SEP_FID_PMod, validated taking into account all 129 SEP events from January 1994 to June 2014, obtained a Mean Absolute Error (MAE) of SEP peak time predictions of 11.3 h, a MAE of peak intensity predictions of 0.53 in log10 units of pfu, and a MAE of SEP end time predictions of 28.8 h. The SEPsFLAREs system also acquires data for solar flares nowcasting (including GSFLAD proxy and SISTED detector from MONITOR’s ESA-funded project; [Hernández-Pajares, M., A. García-Rigo, J.M. Juan, J. Sanz, E. Monte and A. Aragón-Ángel (2012), GNSS measurement of EUV photons flux rate during strong and mid solar flares. Space Weather, Volume 10, Issue 12, doi:10.1029/2012SW000826] and [García-Rigo, A. (2012), Contributions to ionospheric determination with Global Positioning System: solar flare detection and prediction of global maps of Total Electron Content, Ph.D. dissertation. Doctoral Program in Aerospace Science & Technology, Technical University of Catalonia, Barcelona, Spain]).