Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Forecasting Geoffective Events from Solar Wind Data and Evaluating the Most Predictive Features through Machine Learning Approaches

Sabrina Guastavino, Katsiaryna Bahamazava, Emma Perracchione, Fabiana Camattari, Gianluca Audone, Daniele Telloni, Roberto Susino, Gianalfredo Nicolini, Silvano Fineschi, Michele Piana, Anna Maria Massone

TL;DR

The paper tackles forecasting geomagnetic disturbances by forecasting whether the SYM-H index will drop below $-50$ nT one hour ahead using LSTM models trained on in-situ solar wind data from L1 (2005–2019). It introduces magnetic helicity $H_m$ and solar wind energy $E$ as input features and uses a Score-Oriented Loss to handle extreme class imbalance, complemented by a correlation-driven feature ranking to identify the most predictive variables. The authors report strong predictive performance, with mean $TSS$ around $0.87$ and mean $wTSS$ around $0.72$ when including $SYM-H$ as a feature, and demonstrate the consistent importance of energy-related features and helicity. The work has operational relevance for proactive space weather warnings and showcases how derived solar wind quantities can enhance geoeffectiveness forecasting beyond using raw solar wind parameters alone.

Abstract

This study addresses the prediction of geomagnetic disturbances by exploiting machine learning techniques. Specifically, the Long-Short Term Memory recurrent neural network, which is particularly suited for application over long time series, is employed in the analysis of in-situ measurements of solar wind plasma and magnetic field acquired over more than one solar cycle, from $2005$ to $2019$, at the Lagrangian point L$1$. The problem is approached as a binary classification aiming to predict one hour in advance a decrease in the SYM-H geomagnetic activity index below the threshold of $-50$ nT, which is generally regarded as indicative of magnetospheric perturbations. The strong class imbalance issue is tackled by using an appropriate loss function tailored to optimize appropriate skill scores in the training phase of the neural network. Beside classical skill scores, value-weighted skill scores are then employed to evaluate predictions, suitable in the study of problems, such as the one faced here, characterized by strong temporal variability. For the first time, the content of magnetic helicity and energy carried by solar transients, associated with their detection and likelihood of geo-effectiveness, were considered as input features of the network architecture. Their predictive capabilities are demonstrated through a correlation-driven feature selection method to rank the most relevant characteristics involved in the neural network prediction model. The optimal performance of the adopted neural network in properly forecasting the onset of geomagnetic storms, which is a crucial point for giving real warnings in an operational setting, is finally showed.

Forecasting Geoffective Events from Solar Wind Data and Evaluating the Most Predictive Features through Machine Learning Approaches

TL;DR

The paper tackles forecasting geomagnetic disturbances by forecasting whether the SYM-H index will drop below nT one hour ahead using LSTM models trained on in-situ solar wind data from L1 (2005–2019). It introduces magnetic helicity and solar wind energy as input features and uses a Score-Oriented Loss to handle extreme class imbalance, complemented by a correlation-driven feature ranking to identify the most predictive variables. The authors report strong predictive performance, with mean around and mean around when including as a feature, and demonstrate the consistent importance of energy-related features and helicity. The work has operational relevance for proactive space weather warnings and showcases how derived solar wind quantities can enhance geoeffectiveness forecasting beyond using raw solar wind parameters alone.

Abstract

This study addresses the prediction of geomagnetic disturbances by exploiting machine learning techniques. Specifically, the Long-Short Term Memory recurrent neural network, which is particularly suited for application over long time series, is employed in the analysis of in-situ measurements of solar wind plasma and magnetic field acquired over more than one solar cycle, from to , at the Lagrangian point L. The problem is approached as a binary classification aiming to predict one hour in advance a decrease in the SYM-H geomagnetic activity index below the threshold of nT, which is generally regarded as indicative of magnetospheric perturbations. The strong class imbalance issue is tackled by using an appropriate loss function tailored to optimize appropriate skill scores in the training phase of the neural network. Beside classical skill scores, value-weighted skill scores are then employed to evaluate predictions, suitable in the study of problems, such as the one faced here, characterized by strong temporal variability. For the first time, the content of magnetic helicity and energy carried by solar transients, associated with their detection and likelihood of geo-effectiveness, were considered as input features of the network architecture. Their predictive capabilities are demonstrated through a correlation-driven feature selection method to rank the most relevant characteristics involved in the neural network prediction model. The optimal performance of the adopted neural network in properly forecasting the onset of geomagnetic storms, which is a crucial point for giving real warnings in an operational setting, is finally showed.
Paper Structure (8 sections, 7 equations, 2 figures, 3 tables)

This paper contains 8 sections, 7 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Dataset
  3. Prediction and feature ranking
  4. Assessment of results
  5. Recurrent neural network
  6. Correlation-driven feature ranking
  7. Results
  8. Conclusions

Figures (2)

  • Figure 1: Predictions over time on a temporal window of the test set of splitting 1: the top panel represents the prediction when SYM-H is not included in the subset of features, whereas bottom panel represents the one when SYM-H is included between features.
  • Figure 2: Predictions over time on a temporal window of the test set of splitting 1: the top panel represents the prediction when SYM-H is not included in the subset of features, whereas bottom panel represents the one when SYM-H is included between features.