Forecasting the Remaining Duration of an Ongoing Solar Flare

TL;DR

This paper presents a real-time forecasting method using random forest regression to predict the remaining duration of ongoing solar flares based on X-ray light curve data, improving over linear models.

Contribution

The paper introduces a novel real-time prediction approach for solar flare duration using machine learning trained on X-ray data, which is computationally efficient and more accurate than linear regression.

Findings

Random forest outperforms linear regression in predicting flare duration.

The method can be applied in real time during flare events.

Forecasts are based on X-ray light curve features.

Abstract

The solar X-ray irradiance is significantly heightened during the course of a solar flare, which can cause radio blackouts due to ionization of the atoms in the ionosphere. As the duration of a solar flare is not related to the size of that flare, it is not directly clear how long those blackouts can persist. Using a random forest regression model trained on data taken from X-ray light curves, we have developed a direct forecasting method that predicts how long the event will remain above background levels. We test this on a large collection of flares observed with GOES-15, and show that it generally outperforms simple linear regression. This forecast is computationally light enough to be performed in real time, allowing for the prediction to be made during the course of a flare.