An Enhanced LSTM-Piecewise Model for Predicting Ionospheric Scintillation

Muhammad Usama, Kai Guo, Zhipeng Wang, Jifeng Guo

Abstract:	Ionospheric scintillation, affecting global navigation satellite system (GNSS) signals, poses a significant threat to satellite-based communication and navigation systems. This study presents a comprehensive approach to predict the S4 index, a key indicator of ionospheric scintillation intensity, by leveraging a diverse set of space weather and ionospheric parameters. The dataset integrates GNSS-derived S4 and rate of TEC index (ROTI) values with external influencing factors, solar activity parameters, geomagnetic indices, total electron content (TEC) product, and ionosonde-derived characteristics. To model the complex temporal dependencies within this multi-dimensional dataset, an enhanced Bidirectional long short-term memory (BiLSTM) neural network is developed. The model was further optimized using Bayesian optimization to fine-tune its hyperparameters, improving predictive performance and reducing overfitting risks. The model’s performance is then evaluated and predicted S4 values are statistically compared with real measurements using the Nakagami distribution to assess the fading characteristics. Additionally, a comparison between the occurrence rates of real and predicted scintillation events validated the model’s reliability over various ionospheric conditions. Feature importance analysis was conducted to identify which parameters most significantly contributed to S4 index prediction. Results indicate that the Bayesian-optimized BiLSTM model effectively captures the nonlinear dependencies between space weather drivers and scintillation behavior. The system is capable of forecasting, providing valuable support for GNSS reliability and sp ace weather monitoring applications. This study demonstrates a novel data-driven approach that combines physical insight by incorporating physically relevant features and machine learning techniques to improve the prediction of ionospheric scintillation, offering a practical tool for space weather-aware system design.
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:	2277 - 2291
Cite this article:	Usama, Muhammad, Guo, Kai, Wang, Zhipeng, Guo, Jifeng, "An Enhanced LSTM-Piecewise Model for Predicting Ionospheric Scintillation," Proceedings of the 38th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2025), Baltimore, Maryland, September 2025, pp. 2277-2291. https://doi.org/10.33012/2025.20289
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In