Speaker
Description
Solar radio bursts are indicators of eruptive events in the solar corona and interplanetary space, but their detection and interpretation from dynamic spectra is complex and requires manual inspection. To support space weather forecasting, an automated classification system based on deep learning (YOLOv5) with ensemble methods was developed and validated on an international dataset covering spectrometers from 49 CALLISTO stations.
The system identifies and classifies Type II, III, and IV solar radio bursts with an F1 score of 0.74 and a precision of 0.82, demonstrating consistent performance across instruments with varying frequency ranges and noise characteristics. By automating detection and classification, it significantly reduces the time currently required to process radio spectrograms, making this data accessible to forecasters by creating real-time alerts. The e-CALLISTO network provides nearly continuous 24/7 solar monitoring due to its globally distributed receivers.
This tool can be integrated into existing space weather infrastructures (e.g., SIDC space weather centre) to support operational decision-making. Updated every 15 minutes, it delivers near-real-time information on solar radio activity, providing forecasters with complementary data that can be integrated alongside other observations and improving the speed and reliability of space weather services.
