|
Activity Number:
|
81
- Statistical and Machine Learning Efforts on Solar Flare Predictions I
|
|
Type:
|
Topic-Contributed
|
|
Date/Time:
|
Monday, August 9, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
|
Abstract #317599
|
|
|
Title:
|
The Statistical Challenges of Solar Flare Forecasting (and) the Discriminant Analysis Flare Forecasting System at NWRA
|
|
Author(s):
|
KD Leka* and Graham Barnes
|
|
Companies:
|
NorthWest Research Associates and NorthWest Research Associates
|
|
Keywords:
|
Sun;
Probabilistic Forecasting;
NonParametric Discriminant Analysis;
Validation Statistics;
solar flares;
Astrostatistics
|
|
Abstract:
|
The task of predicting physical processes can be limited by our ability to understand the underlying physical system; this challenge is exacerbated when faced with a stochastic process in a remotely-sensed system. Solar flares comprise impactful, sudden enhancements in the radiative output of the Sun with Earth-impacting consequences. The physics of the cause of solar flares we can only investigate by remote observations with no in-situ verification, and through the statistical analysis of disparate and often incomplete datasets which nonetheless cover decades of solar astrophysics. In this talk I will introduce solar flare forecasting basics with an emphasis on the statistical challenges of forecasting rare events for human operators with differing interests, and the topic of statistically evaluating how well forecasts do. I will describe NorthWest Research Associates' implementation of non-parametric discriminant analysis for a flexible statistical classifier infrastructure ("NCI") and a near-real-time forecasting system ("DAFFS"). I will subsequently present some of the key challenges presently facing further progress in this topic.
|
Authors who are presenting talks have a * after their name.
