Hybrid and Comparative Approaches to Space Weather Prediction: Numerical Models and Machine Learning

Space weather involves solar-terrestrial environmental dynamics that impact technology and human activity. Originating above the solar photosphere, the propagation of energy and plasma follows a complex causal chain that spans the interplanetary medium, magnetosphere, ionosphere, upper-atmosphere, and down to the ground. The inherent complexity and multi-scale dynamics of these processes pose significant risks to satellite operations, manned space missions, GNSS, radio communication, and power grids. Sparse observations and growing impacts of space weather highlight the necessity for reliable forecasting tools and deeper physical insight. Physics-based numerical models serve as vital “numerical laboratories” to validate physical hypotheses and complement localized measurements. These models enable the transition from empirical to robust numerical space weather prediction. Concurrently, machine learning (ML) advances space weather research through improved pattern recognition, data assimilation, and model optimization. Synergizing physics and data-driven approaches remains challenging, however, specifically regarding model generalizability and accuracy to predict extreme events.This Research Topic aims to provide a platform for state-of-the-art developments in numerical modeling and machine learning applications within the realm of space weather, promoting research that enhances our understanding of critical physical processes and improves forecasting capabilities for space weather variations. We encourage submissions that rigorously evaluate the comparative strengths and weaknesses of numerical and ML-based modeling approaches. Furthermore, we invite pioneering efforts in the development of "gray-box" models—hybrid approaches that integrate first-principles physics with the data-driven power of machine learning. Such models leverage theoretical knowledge while employing ML to capture intricate system dynamics that remain elusive to pure theory. By highlighting innovations in both standalone and integrated methodologies, this Research Topic seeks to catalyze the transition toward more accurate and reliable space weather prediction.We welcome original research, perspectives, and reviews addressing all facets of numerical modeling and machine learning within the space weather domain. Topics of interest include, but are not limited to:• Development and validation of physics-based numerical models for the Sun-Earth system.• Advances in numerical techniques that enhance the stability, accuracy, and efficiency of the space weather model.• Modeling of specific space weather effects (e.g., single-event effects, deep charging, or atmospheric drag).• Hybrid “gray-box” modeling integrating physical constraints and machine learning• Machine learning applications in space weather forecasting.• Data assimilation techniques that are tailored for space weather applications.• Uncertainty quantification and ensemble modeling approaches.• Comparative studies of empirical, numerical, and ML-based predictive models. Topic editors: Huichao Li, Haopeng Wang, Yi Yang Source: https://www.frontiersin.org/research-topics/77242/hybrid-and-comparative-approaches-to-space-weather-prediction-numerical-models-and-machine-learning