Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cosmic-Ray-Constrained LSTM Model for Geomagnetic Storm Prediction

Zongyuan Ge, Chenwaner Zhang, Wei Zhou, Hongyu Zeng, Guiping Zhou

TL;DR

This work develops a physics-informed LSTM for geomagnetic storm prediction by explicitly incorporating cosmic-ray modulation from neutron monitors as a predictive precursor. Leveraging 1995–2020 multi-source data at 1-hour cadence, the model fuses solar-wind/IMF inputs with cosmic-ray background, transient anomalies, and inter-station coherence to extend lead times and improve 48-hour event-detection performance, achieving RMSEs from $5.106$ to $14.788$ nT and a $25.84\%$ relative gain in F1-score over a purely solar-wind-based model. A targeted optimization reduces input dimensionality and tunes the history window, with a single-layer, 50-unit LSTM yielding the best generalization and a runtime suitable for near-real-time forecasting. The findings support the value of physics-informed ML for space-weather, and point to future directions in multi-task learning, uncertainty quantification, and physics-constraint integration to further enhance reliability and interpretability.

Abstract

Geomagnetic storms (GSTs) driven by solar wind-magnetosphere coupling can severely disrupt technological systems, motivating the need for improved prediction accuracy and longer warning times. In this study, we develop a physics-informed Long Short-Term Memory (LSTM) model that incorporates cosmic-ray flux modulation as an additional precursor signal. As coronal mass ejection (CME)-driven disturbances propagate through the heliosphere, enhanced turbulence and magnetic-field compression reduce galactic cosmic-ray (GCR) flux measured by ground-based neutron monitors (Forbush decreases), providing early information that can precede near-Earth solar-wind signatures by 1--3 days. We integrate multi-source space-weather data, spanning 1995-2020, including cosmic-ray observations, solar wind plasma parameters, interplanetary magnetic-field data, and geomagnetic indices. Based on these data, we construct a 19-dimensional feature vector that includes flux background levels, decrease-related indicators, and inter-station correlation measures as model inputs. Employing a 50-unit LSTM architecture, the proposed model achieves root-mean-square errors (RMSE) of 5.106 nT, 8.315 nT, 10.854 nT, 12.883 nT, and 14.788 nT for 2-, 6-, 12-, 24-, and 48-hour predictions, respectively. Incorporating cosmic-ray information further improves 48-hour forecast skill by up to 25.84% (from 0.178 to 0.224). These results demonstrate the value of physics-informed deep learning and cosmic-ray precursors for advancing space-weather forecasting.

Cosmic-Ray-Constrained LSTM Model for Geomagnetic Storm Prediction

TL;DR

This work develops a physics-informed LSTM for geomagnetic storm prediction by explicitly incorporating cosmic-ray modulation from neutron monitors as a predictive precursor. Leveraging 1995–2020 multi-source data at 1-hour cadence, the model fuses solar-wind/IMF inputs with cosmic-ray background, transient anomalies, and inter-station coherence to extend lead times and improve 48-hour event-detection performance, achieving RMSEs from to nT and a relative gain in F1-score over a purely solar-wind-based model. A targeted optimization reduces input dimensionality and tunes the history window, with a single-layer, 50-unit LSTM yielding the best generalization and a runtime suitable for near-real-time forecasting. The findings support the value of physics-informed ML for space-weather, and point to future directions in multi-task learning, uncertainty quantification, and physics-constraint integration to further enhance reliability and interpretability.

Abstract

Geomagnetic storms (GSTs) driven by solar wind-magnetosphere coupling can severely disrupt technological systems, motivating the need for improved prediction accuracy and longer warning times. In this study, we develop a physics-informed Long Short-Term Memory (LSTM) model that incorporates cosmic-ray flux modulation as an additional precursor signal. As coronal mass ejection (CME)-driven disturbances propagate through the heliosphere, enhanced turbulence and magnetic-field compression reduce galactic cosmic-ray (GCR) flux measured by ground-based neutron monitors (Forbush decreases), providing early information that can precede near-Earth solar-wind signatures by 1--3 days. We integrate multi-source space-weather data, spanning 1995-2020, including cosmic-ray observations, solar wind plasma parameters, interplanetary magnetic-field data, and geomagnetic indices. Based on these data, we construct a 19-dimensional feature vector that includes flux background levels, decrease-related indicators, and inter-station correlation measures as model inputs. Employing a 50-unit LSTM architecture, the proposed model achieves root-mean-square errors (RMSE) of 5.106 nT, 8.315 nT, 10.854 nT, 12.883 nT, and 14.788 nT for 2-, 6-, 12-, 24-, and 48-hour predictions, respectively. Incorporating cosmic-ray information further improves 48-hour forecast skill by up to 25.84% (from 0.178 to 0.224). These results demonstrate the value of physics-informed deep learning and cosmic-ray precursors for advancing space-weather forecasting.
Paper Structure (19 sections, 12 equations, 15 figures)

This paper contains 19 sections, 12 equations, 15 figures.

Table of Contents

  1. Introduction
  2. Data Sources and Preprocessing
  3. Data Correction and Preprocessing
  4. Modeling and Results
  5. Prediction Performance Evaluation
  6. Advantages of the Cosmic-Ray-Constrained Model
  7. Model Optimization
  8. Model Validation
  9. LSTM Neural Network Architecture
  10. Cosmic-Ray Modulation as Feature Input
  11. Discussion
  12. Conclusions
  13. Extended Prediction Results
  14. 2-Hour Prediction Results
  15. 12-Hour Prediction Results
  16. ...and 4 more sections

Figures (15)

  • Figure 1: Comparison of the 6 h-ahead Dst forecasts with observations for the test period.
  • Figure 2: 48 hr Dst index forecasts for 2017--2018 with and without cosmic-ray modulation features. Left panel: with cosmic-ray features. Right panel: without cosmic-ray features. The model with cosmic-ray constraints exhibits improved performance during intense storm intervals.
  • Figure 3: Permutation Importance of input features for Dst Prediction
  • Figure 4: Feature Correlation Heatmap used to diagnose multicollinearity in the input space.
  • Figure 5: LSTM model RMSE across five cross-validation folds (48-hour horizon).
  • ...and 10 more figures