Deep Learning Based Detection of EPBs in GOLD Airglow Images Towards GNSS-RO Back Propagation Validation

Carles Quilis Alfonso, Saleh Javadi, Vinícius Ludwig-Barbosa

Peer Reviewed

Abstract:	This study investigates the potential of machine learning techniques for detecting Equatorial Plasma Bubbles (EPBs) using nighttime airglow imagery from the Global-scale Observations of the Limb and Disk (GOLD) mission. EPBs are ionospheric irregularities characterized by significant plasma density depletions that disrupt trans-ionospheric radio wave propagation, affecting satellite navigation and communication systems. We propose and implement a convolutional encoder-decoder neural network specifically designed for precise pixel-level segmentation of EPB structures within GOLD’s 135.6 nm radiance images. The convolutional neural network demonstrated remarkable performance, achieving high precision and recall, successfully detecting prominent EPBs as well as subtle features overlooked in manual annotations. Results also reveal the network’s capability to generalize beyond explicitly labeled data, indicating its robustness in capturing intricate EPB morphologies. Additionally, a preliminary cross-validation was conducted using GNSS Radio Occultation (RO) data, which showed promising correspondence with the EPB locations identified by the machine learning algorithm. This supports the value of integrating deep learning methods with GNSS-RO techniques to achieve comprehensive global detection and validation of EPBs.
Published in:	Proceedings of the 38th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2025) September 8 - 12, 2025 Hilton Baltimore Inner Harbor Baltimore, Maryland
Pages:	3369 - 3376
Cite this article:	Alfonso, Carles Quilis, Javadi, Saleh, Ludwig-Barbosa, Vinícius, "Deep Learning Based Detection of EPBs in GOLD Airglow Images Towards GNSS-RO Back Propagation Validation," Proceedings of the 38th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2025), Baltimore, Maryland, September 2025, pp. 3369-3376. https://doi.org/10.33012/2025.20392
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In