SPRINTS: A framework for solar-driven event forecasting and research

摘要

Capabilities to predict onset and time profiles of solar-driven events, including solar X-ray flares; solar energetic particles (SEP); coronal mass ejections; and high-speed streams, are critical in mitigating their potential impacts. We introduce the Space Radiation Intelligence System (SPRINTS). This NASA-invested technology integrates preeruptive metadata and forecasts from the MAG4 system with posteruptive metadata in order to produce high fidelity and preeruptive to posteruptive transitional forecasts for solar-driven events. To catalog start and end times of the four solar-driven events, SPRINTS is capable of generating posteruptive forecasts based on automatic detections employed on 30+ years of GOES X-ray and particle data as well as 20+ years of ACE and DSCOVR solar wind data. The prediction results for 1986–2016 presented here are from the SPRINTS posteruptive capability for forecasting SEPs leveraging GOES X-ray metadata. We present onset, peak flux, and time profile SEP forecast metrics and results based on expert-guided, statistical, and machine-learned decision tree models. Operating on data from a 20 year period, a machine-learned decision tree model provided the best results for predicting an S1 event (on the NOAA Space Weather Prediction Center Solar Radiation Scale): 86% probability of detection and 37% false alarm rate. Five flare-related metadata sets were leveraged in the decision tree modeling. Consistently, flare-integrated flux, flare heliolongitude, and flare decay-phase duration were found to be the top three forecasting parameters, while flare magnitude and flare latitude had little to no impact on the SEP forecast model. For the solar-driven events of March 2012, we demonstrate SPRINTS abilities to forecast solar flares, SEP onset, and SEP evolution.