Detection of Concentric Traveling Ionospheric Disturbances in CONUS using Advanced YOLO model

Concentric traveling ionospheric disturbances (cTIDs) represent a distinct class of medium-scale TIDs. They are characterized by the quasi-circular wavefronts in differential total electron content (dTEC) maps. These structures are typically attributed to upward-propagating gravity waves generated by deep convection and severe weather systems. However, cTIDs remain under investigated due to their sparse occurrence and reliance on manual inspection. We construct a manually labeled cTID dataset using a year-long (2023) GNSS-derived dTEC dataset over the region of continental United States (CONUS). Two advanced object detection models, Faster Region-based CNN (Faster R-CNN) and You-Only-Look-Once (YOLO) version 11x (YOLO11x), are developed for automated cTID recognition. Our results demonstrate that the YOLO11x consistently outperforms Faster R-CNN across all CV test datasets, with significant relative improvements for precision, recall, and F1-score. This automatic detection work lays a strong foundation for the further development of automatic characterization of cTID features.