Geomagnetic storm forecasting from solar coronal holes

TL;DR

This paper presents a method to forecast geomagnetic storms caused by solar coronal holes using solar observations and Gaussian Process models, achieving moderate prediction accuracy and extending lead times for space weather forecasting.

Contribution

The study introduces a novel approach to predict geomagnetic storm strength directly from solar coronal hole properties using machine learning, improving space weather prediction capabilities.

Findings

Correlation coefficient for Dst index prediction: R=0.63/0.73

Correlation coefficient for Kp index prediction: R=0.65/0.67

Polarity of IMF at L1 preserved in 83% of cases

Abstract

Coronal holes (CHs) are the source of high-speed streams (HSSs) in the solar wind, whose interaction with the slow solar wind creates corotating interaction regions (CIRs) in the heliosphere. Whenever the CIRs hit the Earth, they can cause geomagnetic storms. We develop a method to predict the strength of CIR/HSS-driven geomagnetic storms directly from solar observations using the CH areas and associated magnetic field polarity. First, we build a dataset comprising the properties of CHs on the Sun, the associated HSSs, CIRs, and orientation of the interplanetary magnetic field (IMF) at L1, and the strength of the associated geomagnetic storms by the geomagnetic indices Dst and Kp. Then, we predict the Dst and Kp indices using a Gaussian Process model, which accounts for the annual variation of the orientation of Earth's magnetic field axis. We demonstrate that the polarity of the IMF at L1 associated with CIRs is preserved in around 83% of cases when compared to the polarity of their CH sources. Testing our model over the period 2010-2020, we obtained a correlation coefficient between the predicted and observed Dst index of R = 0.63/0.73, and Kp index of R = 0.65/0.67, for HSSs having a polarity towards/away from the Sun. These findings demonstrate the possibility of predicting CIR/HSS-driven geomagnetic storms directly from solar observations and extending the forecasting lead time up to several days, which is relevant for enhancing space weather predictions.