Highlights

– an overview of the scientific highlights of OPS2 and SWR6 sessions –

Enno Müller,
Institute for Atrophysics and Geophysics,
G.-A. University of Göttingen

Advancements in space weather research are increasingly propelled by innovative techniques that utilise the power of artificial intelligence and the collective efforts of global communities. At the ESWW 2024, two sessions—Applications of Artificial Intelligence to Space Weather and Space Climate (SWR6) and Severe Space Weather Events and Impacts of May 2024 (OPS2)—exemplified this progressive shift. These sessions highlight the potential of machine learning and citizen science in deepening our understanding of space weather phenomena.

Harnessing Machine Learning for Generative Magnetogram Forecasting
A key highlight featuring the application of machine learning to space weather is the research conducted by Francesco Pio Ramunnoet al (talk on Wednesday, 06.11 at 09:00; contributed highlight text, here). They introduced a cutting-edge approach for forecasting solar magnetograms, an essential tool for understanding the Sun’s magnetic behaviour, which can lead to solar flares and other space weather phenomena that affect our planet. The study uses an advanced artificial intelligence technique known as Denoising Diffusion Probabilistic Models (DDPMs) to predict the evolution of solar magnetic fields 24 hours in advance. This model has shown significant improvements over traditional forecasting methods, particularly in predicting critical features like magnetic flux and the size of active regions.

The breakthrough here is not only in forecasting accuracy but also in the innovative way that machine learning metrics were combined with physical properties of the solar magnetic field, like magnetic flux and active region size. This integrated approach means we can make more reliable predictions about solar activity, ultimately helping us better prepare for potentially disruptive space weather events. The research team is already looking forward to enhancing this model by incorporating more complex data, such as solar velocity fields, which could further improve prediction accuracy. 

Towards Interpretable AI in Ionospheric Modelling

Alan Wood (talk on Tuesday, 05.11 at 10:00) presents another innovative approach that blends statistical modelling with artificial intelligence to improve our understanding of ionospheric variability. In their study, the group utilized a two-step modeling process—first employing Generalized Linear Models (GLM) to determine key factors, and then refining these models with a neural network. This approach retains the interpretability of traditional methods while gaining the accuracy of machine learning.

The study focuses on modeling the behavior of plasma density and thermospheric density in Low Earth Orbit (LEO), utilizing data from ESA’s Swarm mission. By combining these two modeling techniques, the team was able to create models that are not only highly accurate but also transparent in terms of why certain features were included. Such advancements are essential for building trust in AI-driven space weather models, particularly for operational use, where understanding the reasoning behind predictions is as important as the predictions themselves.

The Superstorm of May 2024: Citizen Science in Action

Space weather enthusiasts will recall the exceptional geomagnetic storm that began on May 10, 2024, marking one of the most extreme events of the past two decades. This storm, sparked by a series of coronal mass ejections (CMEs) from active region AR 3664, brought spectacular auroral displays much further south than usual, even reaching regions like Europe, the U.S., and parts of Africa.

What made this event truly special was the role of citizen scientists. With the widespread availability of smartphones and social media, people across the world contributed to tracking the auroras, sharing photos, and documenting their experiences. This unprecedented influx of data allowed the “Auroral Research Coordination: Towards Internationalised Citizen Science” (ARTICS) working group to map the extent of the auroral displays more comprehensively than ever before, highlighting the important contributions of citizen science in areas with limited scientific instrumentation. As Maxime Grandin et al. (talk on Thursday,07.11 at 09:30; contributed highlight text, here) will present, they leveraged these citizen science observations to conduct a detailed analysis of the auroral event, using data from over 30 countries. Their findings suggest that the auroral oval expanded far beyond what existing models had predicted, reaching geomagnetic latitudes between 30° and 60°. Such insights not only advance our understanding of auroral physics but also demonstrate the power of collaborative science, where enthusiastic observers around the world make meaningful contributions to scientific research.