Transfer Learning Classifies Type II and Type III Solar Radio Bursts, Enabling Improved Space Weather Monitoring

Intense bursts of radio emission from the Sun, known as solar radio bursts, pose a risk to radio communications and can signal disruptive space weather events. Herman le Roux, Ruhann Steyn, and colleagues from institutions including the Dublin Institute for Advanced Studies and the Technological University of the Shannon, now present a new automated method for classifying these bursts.

The team successfully applies transfer learning, a technique that leverages pre-trained deep learning models, to distinguish between Type II and Type III solar radio bursts using images of radio spectra. Their work, which evaluated models including VGGnet-19 and YOLOv8, achieves high classification accuracy, with F1 scores ranging from 87% to 92%, and demonstrates a practical solution for classifying these events even with limited data, significantly advancing the potential for real-time monitoring and improved space weather forecasting.

Solar Radio Burst Data Sources Explored The study outlines a range of datasets and data sources pertinent to the classification of solar radio bursts, encompassing both established repositories and those compiled for recent research efforts. Key resources include the CALLISTO Quicklook Solar Spectrogram Plots, data from the Rosse Solar-Terrestrial Observatory, and links provided through the International Space Weather Initiative (ISWI). Additional observations from the NenuFAR facility and the Deimos Solar Radio Spectrometer (DSRT) further enrich the available datasets. Researchers have also contributed datasets derived from the work of Le Roux et al., Scully et al., Wang et al., Wang and Yuan, Zhang et al., and Zhao et al. The primary data format consists of solar spectrograms, which visually represent radio frequency intensity over time and are characterized by parameters such as duration, frequency range, and morphological features. While some datasets, like CALLISTO, are readily accessible, others may require direct contact with the respective researchers for access.

Deep Learning Classifies Solar Radio Burst Types Scientists have developed a methodology for automatically classifying solar radio bursts (SRBs) using deep learning and transfer learning. Recognizing the limited availability of data, particularly for Type II bursts, researchers employed a stratified sampling strategy to create a balanced training dataset. This involved fine-tuning five pre-trained Convolutional Neural Network (CNN) architectures, VGGnet-19, MobileNet, ResNet-152, DenseNet-201, and YOLOv8, using the constructed dataset. This transfer learning approach leverages knowledge from models previously trained on large image datasets, accelerating learning and improving performance with limited SRB data. Researchers froze most layers within each pre-trained network, preserving learned feature extraction, and retrained only the final layers to adapt the models specifically for SRB classification. Testing revealed F1 scores ranging from 87% to 92%, with YOLOv8 achieving the highest performance, demonstrating the effectiveness of transfer learning in automating SRB classification and overcoming data limitations.

Automated Solar Radio Burst Classification via Transfer Learning This work presents a breakthrough in the automated classification of solar radio bursts (SRBs), intense radio emissions from the Sun that can disrupt communications and indicate significant space weather events.

The team constructed a balanced training dataset using spectrogram images from the e-Callisto network, employing a stratified sampling strategy to overcome the limited availability of Type II SRB data. The core of this achievement lies in the application of transfer learning, fine-tuning five established CNN architectures, VGGnet-19, MobileNet, ResNet-152, DenseNet-201, and YOLOv8, to classify the SRBs. This leverages knowledge from pre-trained models, effectively addressing the challenge of limited data and observational variability. Testing revealed consistently high performance across all models, with F1 scores ranging from 87% to 92%, and YOLOv8 demonstrating the highest performance, establishing it as the most effective model for automated SRB classification.

Deep Learning Automates Solar Burst Classification Researchers have developed and tested a series of automated methods for classifying solar radio bursts, intense emissions from the Sun that can disrupt radio communications and indicate larger space weather events. Testing revealed strong performance across all models, with F1 scores ranging from 87% to 92%, demonstrating the potential of deep learning for automated event classification. Notably, the YOLOv8 model consistently outperformed the others, achieving an overall accuracy of 92%. Further research will explore ensemble methods and automated parameter extraction to enable large-scale statistical studies of these solar events. 👉 More information 🗞 Type II and Type III Solar Radio Burst Classification Using Transfer Learning 🧠 ArXiv: https://arxiv.org/abs/2512.11487 Tags: