Deep Learning the Forecast of Galactic Cosmic-Ray Spectra

Abstract

We introduce a novel deep learning framework based on Long Short-Term Memory (LSTM) networks to predict galactic cosmic-ray spectra on a one-day-ahead basis by leveraging historical solar activity data, overcoming limitations inherent in traditional transport models. By flexibly incorporating multiple solar parameters, such as the heliospheric magnetic field, solar wind speed, and sunspot numbers, our model achieves accurate short-term and long-term predictions of cosmic-ray flux. The addition of historical cosmic-ray flux data significantly enhances prediction accuracy, allowing the model to capture complex dependencies between past and future flux variations. Additionally, the model reliably predicts full cosmic-ray spectra for different particle species, enhancing its utility for comprehensive space weather forecasting. Our approach offers a scalable, data-driven alternative to traditional physics-based methods, ensuring robust daily and long-term forecasts. This work opens avenues for advanced models that can integrate broader observational data, with significant implications for space weather monitoring and mission planning.

8 pages, 7 figures, 1 table. Updated version for the publication in ApJL

References in corpus (20)

• Machine learning and the physical sciences
• Machine learning phases of matter
• Solar Modulation of Cosmic Rays
• Deep Neural Networks to Enable Real-time Multimessenger Astrophysics
• Machine Learning in Nuclear Physics
• Machine Learning Phases of Strongly Correlated Fermions
• On the Importance of NNLO QCD and Isospin-breaking Corrections in ${\epsilon }^{\prime }/\epsilon$
• Regressive and generative neural networks for scalar field theory
• A Deep Learning-based Reconstruction of Cosmic Ray-induced Air Showers
• Very local interstellar spectra for galactic electrons, protons and helium
• Heavy quark potential in quark-gluon Plasma: Deep neural network meets lattice quantum chromodynamics
• The behaviour of galactic cosmic ray intensity during solar activity cycle 24
• Global gradients for cosmic-ray protons in the heliosphere during the solar minimum of cycle 23/24
• Investigating the dark matter signal in the cosmic ray antiproton flux with the machine learning method
• deepCR: Cosmic Ray Rejection with Deep Learning
• Inference of cosmic-ray source properties by conditional invertible neural networks
• Deep learning jet modifications in heavy-ion collisions
• Energy spectra of elemental groups of cosmic rays with the KASCADE experiment data and machine learning
• Constraining strongly coupled new physics from cosmic rays with machine learning techniques
• Cosmic-ray energy reconstruction using machine learning techniques