Conveners
CD1 - Combination of physics-based and data-driven methods for space weather forecasting: orals - part 1
- Sabrina Guastavino (Department of Mathematics, University of Genova)
- Dario Del Moro (University of Rome Tor Vergata)
- Kamen Kozarev (Institute of Astronomy and National Astronomical Observatory, Bulgarian Academy of Sciences)
- Filipa S. Barros (LIACC, IRAP, IA)
CD1 - Combination of physics-based and data-driven methods for space weather forecasting: orals - part 2
- Filipa S. Barros (LIACC, IRAP, IA)
- Kamen Kozarev (Institute of Astronomy and National Astronomical Observatory, Bulgarian Academy of Sciences)
- Dario Del Moro (University of Rome Tor Vergata)
- Sabrina Guastavino (Department of Mathematics, University of Genova)
CD1 - Combination of physics-based and data-driven methods for space weather forecasting: Orals - part 3
- Sabrina Guastavino (Department of Mathematics, University of Genova)
- Dario Del Moro (University of Rome Tor Vergata)
- Kamen Kozarev (Institute of Astronomy and National Astronomical Observatory, Bulgarian Academy of Sciences)
- Filipa S. Barros (LIACC, IRAP, IA)
CD1 - Combination of physics-based and data-driven methods for space weather forecasting: Orals - part 4
- Dario Del Moro (University of Rome Tor Vergata)
- Filipa S. Barros (LIACC, IRAP, IA)
- Kamen Kozarev (Institute of Astronomy and National Astronomical Observatory, Bulgarian Academy of Sciences)
- Sabrina Guastavino (Department of Mathematics, University of Genova)
Description
Space weather forecasting can rely on either physics-based or data-driven approaches. On the one hand, physics-based methodologies have deeper historical roots, with physical equations being studied and applied to model solar events and better understand unknown physical processes. On the other hand, data-driven approaches and, specifically, artificial intelligence (AI) algorithms process multi-modal data to identify patterns/correlations with no (or little) reference to physical models.
However, it has been recently explored the possibility to combine both approaches, by leveraging physics to inform the machine learning methods, and applying machine learning to better estimate key parameters in MHD deterministic equations.
This session aims to provide a platform for sharing and discussing research on data-driven and hybrid approaches combining physics-based and AI methodologies in space weather studies, with a focus on forecasting applications. Topics include predicting solar phenomena driving space weather, such as solar flares, coronal mass ejections (CMEs), and Solar Energetic Particles (SEPs), as well as modeling CME and SEP propagation to estimate arrival times at Earth, and predicting geomagnetic disturbances.
Additionally, submissions on space weather-related forecasting applications are encouraged, such as identifying and classifying active regions and detecting solar structures.
As AI techniques have reached a high level of maturity, and recent studies have demonstrated that combining AI with physics-based approaches holds great promise offering reliable tools for space weather forecasting, coupled with the fact that solar activity is currently at its peak (when eruptive phenomena are more frequent and intense) the topic of the proposed session is particularly timely.
Timely and accurate forecasting of interplanetary coronal mass ejections (ICMEs) is essential for mitigating their impact on space- and ground-based infrastructure. While significant advances have been made in predicting ICME arrival times and identifying their in situ signatures, integrating these steps into a continuous operational pipeline remains a challenge.
In this work, we...
Radiation-belt enhancement events, during which electrons reach energies high enough to penetrate spacecraft shielding, pose serious hazards to satellites. Reliable forecasts of both the peak flux level and the duration above an operationally safe threshold would be invaluable to satellite operators. In this talk, I present an algorithm based on Gaussian Processes (GP) to produce probabilistic...
In recent years, studies have shown that it is possible to predict the geoeffectiveness of solar activity (L1 solar wind speed, geomagnetic indices) directly from EUV solar images using deep learning models (Upendran et al. 2020; Bernoux et al. 2022; Brown et al. 2022; Hu et al. 2022; Wang et al. 2025). These models, which are ultimately intended to be used operationally to provide early...
Machine learning (ML) has shown promise in space weather applications, yet its
predictive power is often limited by the scarcity of rare event data and the lack of
physical constraints. In this study, we explore a physics-informed neural network
(PINN) approach that integrates the VERB-CS model with a neural network model
to estimate cold plasma electron density in the plasmasphere. The...
Operational solar flare forecasting requires computationally efficient and energy-optimal methods that maximize the use of available observational resources to deliver timely and reliable predictions. Synoptic full-disk observations from the Solar Dynamics Observatory (SDO) provide continuous monitoring of solar magnetic activity over more than one solar cycle, enabling detailed studies of...
Solar coronal magnetic fields store the magnetic energy that drives solar eruptions, such as flares and coronal mass ejections, which significantly impact space weather. Nonlinear force-free fields (NLFFFs) are commonly used to model the 3D coronal magnetic fields. We develop a physics-informed neural operator (PINO) model that learns the solution operator mapping 2D photospheric vector...
The spatial extension of active regions of the Sun and their associated images can strongly vary from one case to the next. This inhomogeneity is problematic when studying solar flares with convolutionnal neural networks (CNNs) due to their fixed input size. Several processes can be performed to produce a database with homogeneous-sized data, such as coarse resizing, cropping, or padding of...
We investigate data-driven strategies for identifying and predicting geoeffective events using long-term space environment observations. The study explores different unsupervised learning approaches for detecting statistical anomalies in solar wind in-situ measurements and geomagnetic data, with the aim of enhancing our understanding of solar-terrestrial interaction. Such anomalies may...
Abstract
Prediction of total solar irradiance has experienced significant growth and has become an important field of study in solar physics, space climate, and space weather that employs machine learning (ML) as a tool to overcome the challenges of the nonlinear and dynamic nature of solar activities and intricate spatiotemporal relationships in solar irradiance forecasting.
Important...
In this work, we employ an attention-based deep learning approach to predict flare occurrence from multivariate time series of SHARP magnetogram features. The model takes as input active region data over varying time windows and outputs probabilistic predictions for C+-, M+-, or X+-class flare events. To capture the temporal evolution of active regions, the architecture leverages...
Forecasting and understanding space weather remains a fundamental challenge due to the inherently multi-scale nature of plasma dynamics in the Sun-Earth connection. Traditional first-principles models like fully kinetic Particle-in-Cell (PIC) simulations are highly accurate but computationally prohibitive for operational or ensemble forecasting. In this work, we present a novel hybrid...
A series of PCA-based models were previously developed to forecast the total electron content (TEC) variations caused by space weather. The earlier versions used linear regression models to build a forecast, which later was replaced by neural networks (NN). Such models were tested on the TEC data obtained for a European mid-latitudinal region (Iberian Peninsula).
In this work we present a...
As human space activities, such as the Artemis program, become increasingly ambitious, ensuring safety in the space environment has become more critical than ever. In response, Fujitsu Ltd. and Nagoya University are conducting joint research on space weather for future lunar and deep-space exploration. A key focus of this research is solar energetic particle (SEP) events, which are mostly...
The largest solar flares, of class X and above, are associated with strong energetic particle acceleration. The reconnection process thought to be responsible for solar flares can be mimicked with so-called cellular automata. In particular, sandpile models have proven to well reproduce solar flare statistics (Charbonneau et al. 2001) and have recently been shown to be consistent with MHD...
Here we report on a methodology to predict the duration and intensity of space weather events using re-analysis of long-term space weather data to develop a computationally inexpensive algorithm fed only by operationally-available, real-time data streams. As a use case, we apply this methodology to relativistic electron events based on 30+ years of NOAA GOES >2 MeV electron fluxes but the...
Geomagnetic storms are large disruptions of the magnetosphere. These events can interfere with satellites, communication systems, and power grids, causing significant technological and economic damage. Current forecasting models utilise L1 satellite data, constraining lead time to a few hours', often insufficient for effective mitigation. Accurate long-lead forecasts would help protect...
Accurate three-dimensional (3D) characterization of coronal mass ejections (CMEs) is essential for modelling their propagation through interplanetary space and forecasting their arrival time at Earth. However, forecasting accuracy, assessed through platforms such as the Community Coordinated Modeling Center (CCMC) CME Scoreboard, has shown minimal improvement over the past decade, with...
Timely and accurate solar flare forecasting is vital for minimizing the adverse effects of space weather on Earth and in space environments. We present a deep learning framework that integrates multi-modal solar observations—line-of-sight (LoS) magnetograms, continuum intensity images, and EUV observations (171Å, 193Å, and 304Å from SDO/AIA)—alongside physical parameters derived from SHARP...
Understanding how the thermosphere responds to solar activity remains a critical challenge for space situational awareness, with growing relevance as atmospheric heating poses increasing risks to an expanding population of spacecraft and space debris.
We analyze several months of high-resolution orbit decay data from seven satellites spanning altitudes of ~470-810 km, integrated with...
The key challenges in low Earth orbit (LEO) with space operations are tracking and catalogue maintenance for resident space objects (RSOs) including lethal non trackable (LNT) objects, collision and conjunction analysis, manoeuvre planning, re-entry prediction, etc. All these aspects are deeply dependent on drag, which, in turn, has its major source of uncertainty in the thermospheric density....
Solar flares pose risks to infrastructure both on Earth and in space, from induced currents in power grids to satellite damages. Most operational flare forecasting models treat the problem as a binary classification task (flare vs. non-flare) based on a fixed prediction horizon, e.g., 24 hours.
We recast solar-flare forecasting as a continuous time-to-event problem, without the necessity...
Geomagnetic activity indices, such as the well-known Kp index and the recently developed Hpo index, are essential for monitoring and forecasting space weather. These indices provide a global assessment of geomagnetic disturbances caused by solar activity based on data collected from a network of worldwide distributed geomagnetic observatories. Accurate forecasting of these indices is crucial...
Coronal mass ejections (CMEs) are large-scale eruptions of plasma and magnetic flux from the Sun’s corona that propagate through the heliosphere. They play a crucial role in driving space weather phenomena and are responsible for some of the most intense geomagnetic storms. Upon entering the interplanetary space, they are termed interplanetary coronal mass ejections (ICMEs). In-situ...
Geomagnetic storms, characterized by sudden disturbances in Earth's magnetic field, pose significant risks to technological systems and human activities machine learning (ML). Accurate forecasting of geomagnetic storm levels moderate, intense, and super critical for mitigating these impacts. This study assesses the performance of techniques in predicting geomagnetic indices, specifically the...
In the recent years, Physics-Informed Neural Networks (PINNs) have been identified as a promising approach for both forward and inverse modeling problems in physical systems. In particular, applied to Radiation Belt modelling, it has been shown that this methodology is able, to some extent, to reconstruct, from in-situ observation of particle fluxes, the diffusion coefficients which drive the...
We propose a novel sparsity enhancement strategy for regression tasks, based on learning a data-adaptive kernel metric, i.e., a shape matrix, through 2-Layered kernel machines [2]. The resulting shape matrix, which defines a Mahalanobis-type deformation of the input space, is then decomposed via Singular Value Decomposition (SVD), allowing us to identify the most informative directions. This...
Ionospheric electron density and electron temperature affect the telecommunication and navigation/surveying systems such as the Global Navigation Satellite Systems (GNSS). In this study, their inference is based on the Nearest Neighbor (NNB) and Radial Basis Function (RBF) regression models. Synthetic data sets are constructed using data from the International Reference Ionosphere (IRI 2020)...
Solar energetic particles (SEPs) accelerated in coronal mass ejection (CME) driven shocks are key contributors to space weather hazards. However, their acceleration and escape mechanisms remain incompletely understood, particularly under realistic coronal conditions. We present a one-dimensional Monte Carlo simulation framework for modeling SEP acceleration using diffusive shock acceleration...
Solar filaments, phenomena in the solar corona, are of significant scientific interest due to their link with violent eruptive events such as coronal mass ejections. This study introduces a comprehensive deep learning framework for the detection, classification, segmentation, and tracking of solar filaments using H$\alpha$ images from the Global Oscillation Network Group data archive. Using...
We present SunSCC, a fully automated system to detect, aggregate, and classify sunspot groups according to the McIntosh scheme using ground-based white light (WL) observations from the USET facility located at the Royal Observatory of Belgium. The sunspot detection uses a Convolutional Neural Network (CNN), trained from segmentation maps obtained with an unsupervised method based on...
Satellite drag predictions continue to be one of the main challenges facing operators of satellites in Low Earth Orbit (LEO). Drag-validated data assimilation (DA) techniques such as IDEA [Sutton 2018] using an ensemble of global circulation models, and Dragster [Pilinski et al. 2016], using an empirical model ensemble, can determine the thermospheric model forcing that is most compatible...
The most widely used method of forecasting the solar wind conditions in near-Earth space is through a coupled modelling framework. This consists of a coronal model for close to the Sun and a heliospheric model for propagating the solar wind out to Earth. The coronal model is initialised using observations of the photospheric magnetic field and beyond this, when modelling the ambient solar...
