GNSS-VTEC prediction based on CNN-GRU neural network model during high solar activities

Area of Science:

• * Ionospheric physics and space weather.
• * Satellite navigation and positioning systems.
• * Electromagnetic wave propagation.

Background:

• * Total electron content (TEC) is a key ionospheric parameter impacting satellite navigation and space weather.
• * Previous TEC prediction models primarily focused on low solar activity periods.
• * Accurate TEC forecasting is crucial for reliable GNSS operations.

Purpose of the Study:

• * To develop and evaluate a novel CNN-GRU model for TEC forecasting during high solar activity.
• * To assess the model's performance against established empirical and AI-based methods.
• * To enhance the accuracy and reliability of space weather predictions.

Main Methods:

• * Integration of Convolutional Neural Network (CNN) for feature extraction and Gated Recurrent Unit (GRU) for time-series prediction.
• * Utilization of Global Navigation Satellite System (GNSS) data from a single receiver in Sanya, China.
• * Comparative analysis against IRI, NeQuick, GRU, and SVM models.

Main Results:

• * The CNN-GRU model achieved a Root Mean Square Error (RMSE) of 4.28 TECU for 1-hour ahead predictions.
• * For 24-hour forecasts, the CNN-GRU model demonstrated a significantly lower average RMSE of 6.94 TECU.
• * The model outperformed IRI, NeQuick2, SVM, and GRU, showing consistent accuracy across various conditions, including geomagnetic storms.

Conclusions:

• * The CNN-GRU model offers a significant advancement in TEC prediction accuracy, especially during high solar activity.
• * This approach provides a more reliable tool for space weather forecasting and mitigating GNSS errors.
• * The model's robust performance highlights the potential of hybrid deep learning architectures in geophysics.